#!/usr/bin/env bash
# plan-groomer.sh — Plane-1 planner for the R720 agent-team (REPORT-ONLY).
#
# Design refs: docs/r720-agent-team-design.md §4 (planner roster: plan-groomer —
# "Drafts a groomed weekly plan INTO the mode-600 report for now (D3); auto-write to
# Notion/Jira is a later toggle once trusted") and §7 Phase 4 ("planner + confluence-doc").
# This mirrors compliance-drift.sh / dependency-cve.sh conventions VERBATIM so the
# coordinator (§5) can drive it identically — BUT its output discipline is different:
# it is REPORT-ONLY, not ALARM-only.
#
# WHAT IT DOES (read-only):
#   Aggregates the actionable items the OTHER Plane-1 checkers already produced into a
#   single prioritized "groomed weekly plan". It does NOT re-scan repos or hit any network:
#   it reads the LATEST per-checker JSON reports under $REPORT_ROOT_BASE/<checker>/<latest-date>/.
#   Sources consumed (each optional — a missing checker is noted, never invented as work):
#     compliance-drift/<date>/compliance-drift.json   (.findings[])
#     dependency-cve/<date>/dependency-cve.json        (.findings[])
#     doc-drift/<date>/doc-drift.json                  (.findings[], if Phase-3 doc-drift exists)
#     confluence-doc/<date>/confluence-doc.json        (.findings[], the Phase-4 sibling)
#   Each finding is normalized to a plan item {repo, severity, source, title, action},
#   DEDUPED (same repo+source+title collapses), grouped by severity then repo, and written
#   into a prioritized plan in the mode-600 report (JSON + human text).
#
# REPORTING (D3 — REPORT-ONLY, the key difference from the ALARM-only checkers):
#   - Writes a per-run JSON + text report under $REPORT_ROOT/<UTC-date>/, mode 600 (umask 077).
#   - Slack: posts NOTHING by default. A groomed plan is a digest, not an alarm — auto-write
#     to Notion/Jira (or a Slack digest) is a later toggle once signal quality is trusted (D3).
#     There is intentionally NO post_slack_alarm() call in the default path; --notify is a
#     future seam left inert here. A clean week (zero items) still writes an (empty) plan.
#   - Reuses the substrate's redact() for the in-report digest string (defense-in-depth).
#
# SUBSTRATE REUSE (lib/sweep_substrate.sh, sourced — bash dynamic scoping):
#   redact   -> mask any secret-shaped value that leaked into an upstream report title.
#   (discover_repos/mirror_repo/post_slack_alarm are intentionally NOT used: plan-groomer
#    neither clones nor alarms — it only reads sibling reports and writes one mode-600 plan.)
#
# CANARY / DRY-RUN (offline, no network, no token):
#   --canary points $REPORT_ROOT_BASE at a fixture set of sample checker reports
#   (checkers/fixtures/plan-groomer/sample-reports/<checker>/<date>/<checker>.json) and asserts
#   the groomed-plan ITEM COUNT equals EXPECTED_PLAN_ITEMS (anti-complacency floor, design §6.4).
#   If aggregation/dedup regresses, the count drifts and the canary FAILS (exit 3). --canary
#   implies --dry-run. Fully offline + deterministic. --dry-run also suppresses the (inert)
#   --notify seam.
#
# SCOPE / SAFETY:
#   Read-only. No network, no token, no clones, no agent_team/ writes, no systemd wiring —
#   that is provisioning (gated). See "PROVISIONING (NOT DONE HERE)" at the bottom.
#
# Exit: 0 = ran (always, report-only); 2 = setup/usage error; 3 = canary assertion FAILED.
set -euo pipefail
export PATH="$HOME/.local/bin:/opt/homebrew/bin:/usr/local/bin:$PATH"

log() { echo "[plan-groomer] $*" >&2; }
die() { echo "[plan-groomer] FATAL: $*" >&2; exit 2; }

# --- Shared substrate ---------------------------------------------------------
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
SUBSTRATE="$HERE/../lib/sweep_substrate.sh"
[ -f "$SUBSTRATE" ] || die "shared substrate not found: $SUBSTRATE"
# shellcheck source=../lib/sweep_substrate.sh
. "$SUBSTRATE"

# --- Config + defaults (env, all optional) ------------------------------------
GH_ORG="${GH_ORG:-Sea-Haven-Industries}"
# Base under which each checker writes its own <checker>/<date>/ report tree (the parent of
# the per-checker REPORT_ROOTs the other checkers default to: $HOME/sweep-reports).
REPORT_ROOT_BASE="${REPORT_ROOT_BASE:-$HOME/sweep-reports}"
# plan-groomer's OWN report tree (separate from the checkers it reads).
REPORT_ROOT="${REPORT_ROOT:-$HOME/sweep-reports/plan-groomer}"
# Which sibling checkers to aggregate (space-separated; missing ones are noted, never invented).
SOURCE_CHECKERS="${SOURCE_CHECKERS:-compliance-drift dependency-cve doc-drift confluence-doc}"

DRY_RUN=0           # --dry-run: suppress the (inert) --notify seam (report still written).
CANARY=0            # --canary: read the fixture report set + assert the known plan-item count.
NOTIFY=0            # --notify: INERT future seam (post the digest somewhere). Off by default (D3).
TARGETS_OVERRIDE="" # --targets "a b": restrict the groomed plan to these repo names only.

usage() {
  cat >&2 <<EOF
plan-groomer.sh — Plane-1 weekly planner (read-only, REPORT-ONLY per D3)

  --canary          aggregate the fixture report set and assert the known plan-item count
                    (implies --dry-run; fully offline smoke test)
  --dry-run         suppress the (inert) --notify seam (the report is still written)
  --notify          INERT future seam (D3): a later toggle to push the digest to Slack/Notion/Jira.
                    Does nothing in this phase; left here to mark the seam.
  --targets "a b"   restrict the groomed plan to these repo names only
  -h|--help         this help

Env: GH_ORG REPORT_ROOT_BASE REPORT_ROOT SOURCE_CHECKERS
EOF
}

while [ $# -gt 0 ]; do
  case "$1" in
    --canary)   CANARY=1; DRY_RUN=1 ;;
    --dry-run)  DRY_RUN=1 ;;
    --notify)   NOTIFY=1 ;;
    --targets)  shift; TARGETS_OVERRIDE="${1:-}" ;;
    -h|--help)  usage; exit 0 ;;
    *) die "unknown arg: $1 (see --help)" ;;
  esac
  shift
done

command -v jq >/dev/null || die "jq is required"

# --- Report dir (mode 600 reports; matches sweep conventions) -----------------
umask 077
UTC_DATE="$(date -u +%Y-%m-%d)"
UTC_STAMP="$(date -u +%Y-%m-%dT%H:%M:%SZ)"

# Canary redirects the SOURCE base at the fixture report set; output stays in a temp area so
# the canary writes nothing under $HOME.
if [ "$CANARY" -eq 1 ]; then
  FIXTURE_ROOT="$HERE/fixtures/plan-groomer"
  [ -d "$FIXTURE_ROOT/sample-reports" ] || die "canary fixture missing: $FIXTURE_ROOT/sample-reports"
  REPORT_ROOT_BASE="$FIXTURE_ROOT/sample-reports"
  CANARY_OUT="$(mktemp -d "${TMPDIR:-/tmp}/plan-groomer-canary.XXXXXX")"
  trap 'rm -rf "$CANARY_OUT"' EXIT
  REPORT_ROOT="$CANARY_OUT/plan-groomer"
  # Pin the source list the fixtures were authored against (deterministic regardless of env).
  SOURCE_CHECKERS="compliance-drift dependency-cve doc-drift confluence-doc"
fi

REPORT_DIR="$REPORT_ROOT/$UTC_DATE"
mkdir -p "$REPORT_DIR"; chmod 700 "$REPORT_ROOT" "$REPORT_DIR" 2>/dev/null || true
# shellcheck disable=SC2034  # named for parity with the ALARM-only checkers' substrate contract
SWEEP_LOG="$REPORT_DIR/plan-groomer.log"
REPORT_JSON="$REPORT_DIR/plan-groomer.json"
REPORT_TXT="$REPORT_DIR/plan-groomer.txt"

log "=== plan-groomer $UTC_STAMP (canary=$CANARY dry_run=$DRY_RUN notify=$NOTIFY) ==="

# ------------------------------------------------------------------------------
# Severity ordering: the jq sort below maps severity->rank; no shell helper needed.
# ------------------------------------------------------------------------------
in_csv() {  # needle space-list -> 0 if present
  local n="$1" list="$2" t; for t in $list; do [ "$t" = "$n" ] && return 0; done; return 1
}

# --- Resolve the LATEST date dir for one checker under $REPORT_ROOT_BASE -------
latest_report_json() {  # checker_name -> path to its latest <date>/<checker>.json, or "" if none
  local checker="$1"
  local base="$REPORT_ROOT_BASE/$checker" d name latest=""
  [ -d "$base" ] || { echo ""; return 0; }
  # Date dirs are YYYY-MM-DD; lexical sort == chronological. Glob the dirs, keep only
  # YYYY-MM-DD names, sort newest-first, pick the newest that actually has the JSON.
  for d in $(for p in "$base"/*/; do
               [ -d "$p" ] || continue
               name="$(basename "$p")"
               [[ "$name" =~ ^[0-9]{4}-[0-9]{2}-[0-9]{2}$ ]] && echo "$name"
             done | sort -r); do
    if [ -f "$base/$d/$checker.json" ]; then latest="$base/$d/$checker.json"; break; fi
  done
  echo "$latest"
}

# ==============================================================================
# AGGREGATE: pull findings from each sibling checker's latest report into plan items.
# ==============================================================================
declare -a PLAN_ITEMS=()      # one normalized JSON plan-item per upstream finding
declare -a MISSING_SOURCES=() # checkers with no report found — noted, NEVER invented as work
note_missing() { MISSING_SOURCES+=( "$1" ); }

for checker in $SOURCE_CHECKERS; do
  rj="$(latest_report_json "$checker")"
  if [ -z "$rj" ]; then
    note_missing "$checker(no-report-found)"
    log "  [$checker] no report under $REPORT_ROOT_BASE/$checker — skipped (not invented as work)"
    continue
  fi
  if ! jq -e '.findings | type=="array"' "$rj" >/dev/null 2>&1; then
    note_missing "$checker(report-unparseable)"
    log "  [$checker] report $rj has no .findings[] array — skipped"
    continue
  fi
  cnt="$(jq '.findings | length' "$rj")"
  log "  [$checker] $rj -> $cnt finding(s)"
  # Normalize each finding to a plan item. Title/severity/repo come straight off the finding
  # schema the checkers emit (see finding.schema.json spirit). The "action" is a stable,
  # source-derived hint (no fabrication — it just labels what kind of remediation this is).
  while IFS= read -r item; do
    [ -n "$item" ] && PLAN_ITEMS+=( "$item" )
  done < <(jq -c --arg src "$checker" '
    .findings[]
    | { repo: (.repo // "unknown"),
        severity: (.severity // "medium"),
        source: $src,
        title: (.title // .id // "untitled"),
        action: ( if $src=="dependency-cve" then "upgrade vulnerable dependency"
                  elif $src=="compliance-drift" then "fix convention drift"
                  elif $src=="doc-drift" then "reconcile docs with code"
                  elif $src=="confluence-doc" then "close documentation gap"
                  else "review finding" end ) }
  ' "$rj")
done

# --- Optional repo-scope restriction (--targets) ------------------------------
if [ -n "$TARGETS_OVERRIDE" ] && [ "${#PLAN_ITEMS[@]}" -gt 0 ]; then
  declare -a FILTERED=()
  for item in "${PLAN_ITEMS[@]}"; do
    r="$(echo "$item" | jq -r '.repo')"
    in_csv "$r" "$TARGETS_OVERRIDE" && FILTERED+=( "$item" )
  done
  PLAN_ITEMS=( "${FILTERED[@]+"${FILTERED[@]}"}" )
  log "targets filter '$TARGETS_OVERRIDE' -> ${#PLAN_ITEMS[@]} item(s)"
fi

# ==============================================================================
# DEDUP + PRIORITIZE: collapse identical (repo|source|title), then sort by severity desc,
# then repo, then source. Add a stable rank int so downstream consumers can re-sort.
# ==============================================================================
if [ "${#PLAN_ITEMS[@]}" -gt 0 ]; then
  RAW_JSON="$(printf '%s\n' "${PLAN_ITEMS[@]}" | jq -cs .)"
else
  RAW_JSON="[]"
fi

GROOMED_JSON="$(echo "$RAW_JSON" | jq -c '
  # dedup on repo|source|title
  ( reduce .[] as $x ({}; .[($x.repo+"|"+$x.source+"|"+$x.title)] //= $x) ) | [ .[] ]
  | map(. + { rank: ( {critical:4, high:3, medium:2, low:1}[.severity] // 0 ) })
  | sort_by([ (-.rank), .repo, .source, .title ])
')"

N_ITEMS="$(echo "$GROOMED_JSON" | jq 'length')"
N_CRIT_HIGH="$(echo "$GROOMED_JSON" | jq '[.[]|select(.severity=="critical" or .severity=="high")]|length')"
N_REPOS="$(echo "$GROOMED_JSON" | jq '[.[].repo]|unique|length')"

if [ "${#MISSING_SOURCES[@]}" -gt 0 ]; then
  MISSING_JSON="$(printf '%s\n' "${MISSING_SOURCES[@]}" | jq -R . | jq -cs .)"
else
  MISSING_JSON="[]"
fi

# ==============================================================================
# ASSEMBLE REPORT (JSON + text), mode 600
# ==============================================================================
jq -n \
  --arg planner "plan-groomer" --arg ts "$UTC_STAMP" --arg org "$GH_ORG" \
  --arg sources "$SOURCE_CHECKERS" \
  --argjson items "$GROOMED_JSON" --argjson missing "$MISSING_JSON" \
  '{planner:$planner, generated:$ts, org:$org, mode:"report-only",
    sources_considered:($sources|split(" ")),
    plan_item_count:($items|length),
    crit_high_count:([$items[]|select(.severity=="critical" or .severity=="high")]|length),
    repos_in_plan:([$items[].repo]|unique|length),
    plan:$items, missing_sources:$missing}' > "$REPORT_JSON"

{
  echo "plan-groomer — groomed weekly plan — $UTC_STAMP"
  echo "org=$GH_ORG sources=[$SOURCE_CHECKERS] mode=report-only (D3: no auto-write)"
  echo "plan items: $N_ITEMS ($N_CRIT_HIGH crit/high) across $N_REPOS repo(s)"
  echo
  if [ "$N_ITEMS" -gt 0 ]; then
    echo "PRIORITIZED PLAN (severity desc, then repo):"
    echo "$GROOMED_JSON" | jq -r '.[] | "• [\(.severity)] \(.repo) — \(.title)\n    action: \(.action) (source: \(.source))"'
  else
    echo "No actionable items aggregated this run (clean week, or no upstream reports)."
  fi
  if [ "$(echo "$MISSING_JSON" | jq 'length')" -gt 0 ]; then
    echo; echo "sources with no report (NOT invented as work):"
    echo "$MISSING_JSON" | jq -r '.[] | "  - \(.)"'
  fi
} > "$REPORT_TXT"
chmod 600 "$REPORT_JSON" "$REPORT_TXT" 2>/dev/null || true

# Defense-in-depth: the digest line that a future --notify seam would push is redacted now.
DIGEST="$(echo "$GROOMED_JSON" | jq -r '
  group_by(.repo)[] | "*\(.[0].repo)*: " + ([.[] | "[\(.severity)] \(.title)"] | join("; "))' \
  | sed 's/^/• /' | redact)"

log "report: $REPORT_JSON ($N_ITEMS plan item(s), $N_REPOS repo(s))"

# ==============================================================================
# CANARY ASSERTION (anti-complacency floor, design §6.4)
# ==============================================================================
if [ "$CANARY" -eq 1 ]; then
  EXPECT_FILE="$HERE/fixtures/plan-groomer/EXPECTED_PLAN_ITEMS"
  [ -f "$EXPECT_FILE" ] || die "canary expected-count file missing: $EXPECT_FILE"
  EXPECTED="$(tr -dc '0-9' < "$EXPECT_FILE")"
  log "canary assertion: expected plan items=$EXPECTED, got=$N_ITEMS"
  if [ "$N_ITEMS" -ne "$EXPECTED" ]; then
    echo "[plan-groomer] CANARY FAIL: groomed-plan item count mismatch (expected $EXPECTED, got $N_ITEMS)" >&2
    echo "  -> aggregation or dedup regressed, or the fixture changed. See $REPORT_TXT." >&2
    exit 3
  fi
  log "canary PASS: groomed plan has all $EXPECTED expected item(s) (dedup intact)."
fi

# ==============================================================================
# REPORT-ONLY ROUTING (D3): the plan lives in the mode-600 report. Post NOTHING.
# ==============================================================================
if [ "$NOTIFY" -eq 1 ] && [ "$DRY_RUN" -eq 0 ]; then
  # INERT future seam: when D3's "once trusted" toggle flips, this is where the digest would
  # be pushed to Slack/Notion/Jira. It is intentionally a no-op in this phase — plan-groomer
  # is REPORT-ONLY and must not auto-write. The composed digest is available in $DIGEST.
  log "--notify requested but inert in this phase (D3: report-only; auto-write is a later toggle). No push."
fi
: "${DIGEST:?}" >/dev/null 2>&1 || true   # DIGEST is composed for the future seam; keep it referenced.
log "REPORT-ONLY: groomed plan written to the mode-600 report; nothing posted (D3)."
exit 0

# ==============================================================================
# PROVISIONING (NOT DONE HERE — gated, later phases):
#   - REPORT-ONLY by design (D3). The auto-write path (push the groomed plan to Notion/Jira,
#     or a weekly Slack digest) is a LATER TOGGLE, flipped only once signal quality is trusted.
#     The --notify seam above is intentionally inert; wiring a real destination is provisioning.
#   - No systemd unit / timer is installed by this script. Wiring it into the weekly schedule
#     (alongside the other Plane-1 checkers under the coordinator) is provisioning and is gated.
#   - The coordinator (design §5, checker_coordinator.sh) registers + drives this planner; that
#     registry edit is done centrally, NOT in this script.
#   - Confluence + project_r720_agent_team memory updates are docs-as-you-go obligations for the
#     build session, tracked outside this script.
# ==============================================================================