diff --git a/docs/upstream-sync/triage.jsonl b/docs/upstream-sync/triage.jsonl
index 071aba9a..41ed31d8 100644
--- a/docs/upstream-sync/triage.jsonl
+++ b/docs/upstream-sync/triage.jsonl
@@ -1,4 +1,4 @@
-{"_meta": {"last_synced": "60e7307c", "last_synced_date": "2026-07-23"}}
+{"_meta": {"last_synced": "0e2eefbc", "last_synced_date": "2026-07-24"}}
 {"sha": "0b76afdc", "pr": 1653, "subject": "reviews block agenda, sticky headers, diff scroll", "disposition": "landed", "reason": "", "branch": "cherry-pick-upstream", "local_sha": null, "updated": "2026-07-02T00:00:00Z"}
 {"sha": "7530653b", "pr": 1655, "subject": "ResizeObserver settle for review scroll-to", "disposition": "landed", "reason": "", "branch": "cherry-pick-upstream", "local_sha": null, "updated": "2026-07-02T00:00:00Z"}
 {"sha": "23bd4a63", "pr": 1660, "subject": "top padding to sticky review block header", "disposition": "landed", "reason": "", "branch": "cherry-pick-upstream", "local_sha": null, "updated": "2026-07-02T00:00:00Z"}
@@ -146,3 +146,7 @@
 {"sha": "bedc57b8", "pr": 1796, "subject": "fix: cap execute output by making SandboxBackendProxy a BaseSandbox (#1796)", "disposition": "deferred", "reason": "Real fix (unbounded execute stdout pulled into the worker → OOM risk) but inapplicable at the fork's deepagents==0.6.12: no capture-offload API exists (ExecuteOffloadResult / execute_accepts_timeout absent; FilesystemMiddleware has no _resolve_capture BaseSandbox gate), so the bug it fixes cannot occur yet. Re-triage together with deferred #1745 (deepagents 0.6.12→0.7.x bump) — landing that bump makes this fix required. Fork's SandboxBackendProxy has diverged (sync-era, bound_repo repo-binding, no reconnect machinery): hand-apply the execute_with_offload/aexecute_with_offload delegation onto the fork proxy rather than cherry-pick.", "branch": "", "local_sha": null, "updated": "2026-07-23T16:54:54Z"}
 {"sha": "e1138cf5", "pr": 1797, "subject": "fix: merge concurrent trusted skill refs (#1797)", "disposition": "wont-merge", "reason": "Follow-up fix to #1774's TrustedSkillsMiddleware, which the fork never adopted (row df743658) and upstream itself reverted in #1800 — nothing to apply.", "branch": "", "local_sha": null, "updated": "2026-07-23T16:54:54Z"}
 {"sha": "60e7307c", "pr": 1800, "subject": "revert: repository skill support for the coding agent (#1774, #1797) (#1800)", "disposition": "wont-merge", "reason": "Revert of #1774 + #1797; the fork adopted neither, so there is nothing to revert. Upstream's rationale: #1774's factory-time ensure_sandbox_for_thread ran outside the serialized run and raced the thread-deterministic sandbox name (prod create-409s). Fork invariant holds: its single provisioning call sits inside the __creating__-sentinel lifecycle within the interrupt-serialized run.", "branch": "", "local_sha": null, "updated": "2026-07-23T16:54:54Z"}
+{"sha": "a77c4e47", "pr": 1799, "subject": "fix: show current shared thread in sidebar (#1799)", "disposition": "deferred", "reason": "Desirable dashboard fix: sidebar now surfaces the currently-open shared thread (org-member 'Open in Web' links) when it falls outside the caller's own list — _sidebar_active_thread_summary + active_thread_id param; fork has the same include_all/org-member read path in thread_api.py. Cherry-picks CLEAN onto dev (merge-tree probe: zero conflicts); port with its thread-api tests + sandbox_id e2e spec + AgentsSidebar/queries UI as one vertical.", "branch": "", "local_sha": null, "updated": "2026-07-24T19:25:32Z"}
+{"sha": "b67215ed", "pr": 1780, "subject": "feat: link originating Slack threads (#1780)", "disposition": "deferred", "reason": "Desirable for the Slack-heavy fork: dispatch records the originating Slack permalink in thread metadata (webhooks/common.py), open_pull_request embeds an 'Originating Slack thread' link in the PR body (gated to private repos), and the sidebar menu links it. Impl files auto-merge (thread_api / open_pull_request / webhooks/common); conflicts are fork drift in prompt.py (customized prompt constants), tests/e2e/harness.py, tests/slack/test_slack_context.py — hand re-key those; keep backend + UI (AgentsSidebar, types) + e2e dashboard spec as one vertical.", "branch": "", "local_sha": null, "updated": "2026-07-24T19:25:37Z"}
+{"sha": "7d79d300", "pr": 1804, "subject": "chore(deps): bump deepagents to 0.7.0b1 pin (#1804)", "disposition": "deferred", "reason": "Rides the deferred #1745 deepagents bump: 0.7.0a7→0.7.0b1 pin + wcmatch>=11.0 uv override (langchain-e2b's e2b dep still caps wcmatch<11.0). Fork is pinned deepagents==0.6.12 — and langsmith==0.10.10 already satisfies the new >=0.10.9 floor — and never took the a7-era HARNESS_EXCLUDED_MIDDLEWARE code its prompt.py hunk deletes, so nothing applies standalone. Re-triage as one cluster: #1745 (bump, now targeting 0.7.0b1) + #1796 (offload delegation) + this pin/override.", "branch": "", "local_sha": null, "updated": "2026-07-24T19:25:44Z"}
+{"sha": "0e2eefbc", "pr": 1801, "subject": "chore: bump default sandbox resources and delete-after-stop (#1801)", "disposition": "deferred", "reason": "Capacity/cost policy, not a correctness fix: defaults 2→4 vCPU, 7936MiB→16GiB mem, 32→128GiB FS, stopped-sandbox delete 24h→14d. Would materially raise LangSmith sandbox spend on sh-openswe unless prod env already overrides; all five values are env-overridable today. Decide prod sizing first (ops decision), then adopt or reject. Textual conflict is only fork drift in integrations/langsmith.py (sync-era SandboxClient, reconnect constants absent) — the constants re-key trivially; docs hunks auto-merge.", "branch": "", "local_sha": null, "updated": "2026-07-24T19:25:49Z"}
diff --git a/docs/upstream-sync/triage.md b/docs/upstream-sync/triage.md
index b56c8b79..9c2cf066 100644
--- a/docs/upstream-sync/triage.md
+++ b/docs/upstream-sync/triage.md
@@ -6,7 +6,7 @@ Commits on `upstream/main` (langchain-ai/open-swe) not yet in `dev`, and the dec
 Rows key on the **upstream SHA** (stable across local cherry-picks). Deferred rows are provisional
 — re-inspect before picking. See the fork-maintenance runbook in `CLAUDE.md`.
 
-**Last synced `upstream/main`:** `60e7307c` (2026-07-23)
+**Last synced `upstream/main`:** `0e2eefbc` (2026-07-24)
 
 | sha | pr | subject | decision | why | branch |
 |---|---|---|---|---|---|
@@ -157,5 +157,9 @@ Rows key on the **upstream SHA** (stable across local cherry-picks). Deferred ro
 | `e51abe14` | #1754 | fix: skip oversized images before model calls (#1754) | Deferred | Desirable: 10MB image size cap + 'image omitted' text block before model calls, prevents oversized-image model failures. Conflicts only because the fork hardened fetch_image_block (SSRF redirect guard, host-only logging) — hunks are compatible; re-key the size check around the fork's guarded fetch and port impl + test together. |  |
 | `0ed560a5` | #1779 | fix: settle review checks after run failures (#1779) | Deferred | Desirable: run-failure completion webhook now settles reviewer check runs left open when the graph dies (_settle_failed_reviewer_check; uses settle_review_check_run + review_check_pending_result, both already in the fork's review/publish.py from the #214/#215 ports). Conflict is fork drift in completion.py (failure-reply customizations); re-key the new helper in and port with its 159-line test. |  |
 | `bedc57b8` | #1796 | fix: cap execute output by making SandboxBackendProxy a BaseSandbox (#1796) | Deferred | Real fix (unbounded execute stdout pulled into the worker → OOM risk) but inapplicable at the fork's deepagents==0.6.12: no capture-offload API exists (ExecuteOffloadResult / execute_accepts_timeout absent; FilesystemMiddleware has no _resolve_capture BaseSandbox gate), so the bug it fixes cannot occur yet. Re-triage together with deferred #1745 (deepagents 0.6.12→0.7.x bump) — landing that bump makes this fix required. Fork's SandboxBackendProxy has diverged (sync-era, bound_repo repo-binding, no reconnect machinery): hand-apply the execute_with_offload/aexecute_with_offload delegation onto the fork proxy rather than cherry-pick. |  |
+| `a77c4e47` | #1799 | fix: show current shared thread in sidebar (#1799) | Deferred | Desirable dashboard fix: sidebar now surfaces the currently-open shared thread (org-member 'Open in Web' links) when it falls outside the caller's own list — _sidebar_active_thread_summary + active_thread_id param; fork has the same include_all/org-member read path in thread_api.py. Cherry-picks CLEAN onto dev (merge-tree probe: zero conflicts); port with its thread-api tests + sandbox_id e2e spec + AgentsSidebar/queries UI as one vertical. |  |
+| `b67215ed` | #1780 | feat: link originating Slack threads (#1780) | Deferred | Desirable for the Slack-heavy fork: dispatch records the originating Slack permalink in thread metadata (webhooks/common.py), open_pull_request embeds an 'Originating Slack thread' link in the PR body (gated to private repos), and the sidebar menu links it. Impl files auto-merge (thread_api / open_pull_request / webhooks/common); conflicts are fork drift in prompt.py (customized prompt constants), tests/e2e/harness.py, tests/slack/test_slack_context.py — hand re-key those; keep backend + UI (AgentsSidebar, types) + e2e dashboard spec as one vertical. |  |
+| `7d79d300` | #1804 | chore(deps): bump deepagents to 0.7.0b1 pin (#1804) | Deferred | Rides the deferred #1745 deepagents bump: 0.7.0a7→0.7.0b1 pin + wcmatch>=11.0 uv override (langchain-e2b's e2b dep still caps wcmatch<11.0). Fork is pinned deepagents==0.6.12 — and langsmith==0.10.10 already satisfies the new >=0.10.9 floor — and never took the a7-era HARNESS_EXCLUDED_MIDDLEWARE code its prompt.py hunk deletes, so nothing applies standalone. Re-triage as one cluster: #1745 (bump, now targeting 0.7.0b1) + #1796 (offload delegation) + this pin/override. |  |
+| `0e2eefbc` | #1801 | chore: bump default sandbox resources and delete-after-stop (#1801) | Deferred | Capacity/cost policy, not a correctness fix: defaults 2→4 vCPU, 7936MiB→16GiB mem, 32→128GiB FS, stopped-sandbox delete 24h→14d. Would materially raise LangSmith sandbox spend on sh-openswe unless prod env already overrides; all five values are env-overridable today. Decide prod sizing first (ops decision), then adopt or reject. Textual conflict is only fork drift in integrations/langsmith.py (sync-era SandboxClient, reconnect constants absent) — the constants re-key trivially; docs hunks auto-merge. |  |
 
 _Maintenance: after a `git sync`, add new `dev..upstream/main` SHAs as **Untriaged** (edit `triage.jsonl`) and bump "Last synced". A successful `git cherry-pick -x` auto-moves the row to **Landed** via the `post-commit` journal + `make triage-reconcile`._