You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
plugins/claude-ops/skills/lanes/context/refresh.md's git staleness probe (added in #514 / PR #791) references a <lane-launch-commit> placeholder described as "the repo HEAD when lanes start/restart last ran" — but plugins/claude-ops/skills/lanes/scripts/lane-launcher.sh never captures this value anywhere (confirmed by inspection: the only rev-parse call in that script is rev-parse --show-toplevel for the repo root). An operator or future automation following the probe literally has no automated way to retrieve <lane-launch-commit> — it's advisory-only today.
Observed vs expected
Observed: the probe is conceptually correct (git log --oneline "<lane-launch-commit>..${default}" -- plugins/claude-ops/) but the placeholder has no producer.
Expected: lane-launcher.sh captures git rev-parse HEAD at launch time (after its pre-launch pull step) and persists it somewhere the probe doc can point at — e.g. ${CLAUDE_PLUGIN_DATA}/lanes/<lane>-launch-commit.
Fix direction (from two independent review passes on PR #791)
Launcher patch: after the pre-launch git pull, capture git rev-parse HEAD and write it to a per-lane file under ${CLAUDE_PLUGIN_DATA}/lanes/. Update context/refresh.md's probe instructions to read from that file instead of an unfilled placeholder. A hex-only SHA from git rev-parse HEAD carries no injection risk; if this is ever instead sourced from external/user-controlled input, that input must never be interpolated unquoted into the probe command — call this out explicitly in the doc once the launcher change lands.
Lighter-weight fallback if the launcher change is out of scope for now: document claude agents --json's startedAt field as a time-based approximation (git log --after="<startedAt>" origin/main -- plugins/claude-ops/).
Category
behavioral (documentation gap / incomplete automation surface, not a bug in shipped behavior — the probe is correctly labeled advisory)
Lane / source
work lane, 2026-07-21. Deferred VALID finding from PR #791's post-green review pass (two independent review-bot passes both flagged it, before and after a follow-up commit; both classified it non-blocking for a doc-only PR).
Triage: verified (VALID deferred review finding, confirmed by inspection) — BRIEFED but BLOCKED (not routed status: ready, per rule 8d). Decision-defaulted.
Category: behavioral (incomplete automation surface / doc gap — the probe is correctly labeled advisory today, not a shipped bug). Priority medium (rule 5 default; claude-ops sits outside the week's priority-plugin set → order low within tier; lane-launcher is pipeline machinery, so pipeline-correctness weight applies to within-tier ordering only, never a priority grant per 8c).
The refresh.md staleness probe this item makes fillable is introduced by PR #791 (closes #514), still OPEN/unmerged. Editing refresh.md's probe now would conflict with that PR. When #514 closes (PR #791 merges), promote this to status: ready — the brief below is already complete.
Problem (behavioral)
claude-ops refresh.md's git-staleness probe references a <lane-launch-commit> placeholder ("the repo HEAD when lanes start/restart last ran"), but the lane launcher never captures that value — its only rev-parse is --show-toplevel for the repo root. The placeholder has no producer, so the probe is unfillable by automation and advisory-only.
Acceptance (behavioral contract)
The lane launcher, after its pre-launch pull, captures the repo HEAD commit and persists it to a per-lane location under the plugin data home (${CLAUDE_PLUGIN_DATA}/lanes/), so each lane's launch commit is retrievable by lane name.
refresh.md's staleness probe reads the launch commit from that persisted location instead of an unfilled placeholder, making the probe runnable end-to-end.
Security note to carry into the doc: a git rev-parse HEAD SHA is hex-only and injection-free; if the launch commit is ever instead sourced from external/user-controlled input, it must never be interpolated unquoted into the probe command — state this explicitly in the doc once the launcher change lands.
Decision defaulted: full launcher-capture fix (recommended) — veto before merge
Two forks: (recommended) the launcher captures + persists HEAD and the probe reads it — the root-cause fix; (alternative) a lighter doc-only stopgap documenting claude agents --json's startedAt as a time-based approximation (git log --after="<startedAt>" origin/main -- plugins/claude-ops/). The stopgap is a workaround, not a fix; recommend the launcher change. A maintainer may veto to the stopgap if the launcher change is out of scope for now.
Source: deferred VALID finding from PR #791's post-green review (two independent review passes both flagged it, both classified non-blocking for a doc-only PR).
Unblocked → promoted to status: ready. Blocker #514 closed (PR #791 merged 2026-07-21T05:04:16Z); the refresh.md staleness probe is now on main, so the brief above is fully workable. Recommended fork stands (launcher captures + persists HEAD; probe reads it) — decision-defaulted, veto before merge.
(Manual promotion by the triage lane — the blocked-by #514 edge auto-resolved on close, but no mechanism auto-flipped this to ready; logged as the re-promotion-gap observation in #657.)
Superseded — this item was already ratified and should not have been queued.
The earlier first-drain ratification request on this issue was raised in error. The admission gate dispatches a C3 bug-fix-shaped item when it carries either its own recorded ratification or a completed first drain — and this issue already carried recorded ratification in its body from attended triage on 2026-07-23. The drain cycle applied the blanket first-drain rule without first reading bodies for an existing ratification, so it queued an item that was already admitted.
Restored to the autonomous frontier: agent-ready re-applied, needs-human removed. No human action is required, and this issue no longer appears in the attended queue's ratify view.
This was generated by AI during work-loop execution.
Autonomous claim confirmed. This item was pre-gated by the work-loop's own admission pass before dispatch (agent-ready + work-class: scoped, no needs-human), so the interactive selection/confirmation prompt in /work-items:work was satisfied by that prior admission decision rather than a user prompt — this comment records that per the skill's autonomous-invocation protocol. Re-verified immediately before claim: no open PR or branch referenced #792. Claimed via the seam (assignee + lease) and now implementing per the item's acceptance criteria in an out-of-tree worktree.
Activity
kyle-sexton commentedon Jul 21, 2026
This was generated by AI during triage.
Triage: verified (VALID deferred review finding, confirmed by inspection) — BRIEFED but BLOCKED (not routed
status: ready, per rule 8d). Decision-defaulted.Category: behavioral (incomplete automation surface / doc gap — the probe is correctly labeled advisory today, not a shipped bug). Priority medium (rule 5 default; claude-ops sits outside the week's priority-plugin set → order low within tier; lane-launcher is pipeline machinery, so pipeline-correctness weight applies to within-tier ordering only, never a priority grant per 8c).
Blocked by #514 — do not pick up yet
The
refresh.mdstaleness probe this item makes fillable is introduced by PR #791 (closes #514), still OPEN/unmerged. Editingrefresh.md's probe now would conflict with that PR. When #514 closes (PR #791 merges), promote this tostatus: ready— the brief below is already complete.Problem (behavioral)
claude-ops
refresh.md's git-staleness probe references a<lane-launch-commit>placeholder ("the repo HEAD whenlanes start/restartlast ran"), but the lane launcher never captures that value — its onlyrev-parseis--show-toplevelfor the repo root. The placeholder has no producer, so the probe is unfillable by automation and advisory-only.Acceptance (behavioral contract)
${CLAUDE_PLUGIN_DATA}/lanes/), so each lane's launch commit is retrievable by lane name.refresh.md's staleness probe reads the launch commit from that persisted location instead of an unfilled placeholder, making the probe runnable end-to-end.git rev-parse HEADSHA is hex-only and injection-free; if the launch commit is ever instead sourced from external/user-controlled input, it must never be interpolated unquoted into the probe command — state this explicitly in the doc once the launcher change lands.Decision defaulted: full launcher-capture fix (recommended) — veto before merge
Two forks: (recommended) the launcher captures + persists HEAD and the probe reads it — the root-cause fix; (alternative) a lighter doc-only stopgap documenting
claude agents --json'sstartedAtas a time-based approximation (git log --after="<startedAt>" origin/main -- plugins/claude-ops/). The stopgap is a workaround, not a fix; recommend the launcher change. A maintainer may veto to the stopgap if the launcher change is out of scope for now.Source: deferred VALID finding from PR #791's post-green review (two independent review passes both flagged it, both classified non-blocking for a doc-only PR).
feat(claude-ops): document lane mid-session staleness & restart caden…
kyle-sexton commentedon Jul 21, 2026
This was generated by AI during triage.
Unblocked → promoted to
status: ready. Blocker #514 closed (PR #791 merged 2026-07-21T05:04:16Z); therefresh.mdstaleness probe is now onmain, so the brief above is fully workable. Recommended fork stands (launcher captures + persists HEAD; probe reads it) — decision-defaulted, veto before merge.(Manual promotion by the triage lane — the
blocked-by #514edge auto-resolved on close, but no mechanism auto-flipped this to ready; logged as the re-promotion-gap observation in #657.)kyle-sexton commentedon Jul 24, 2026
This was generated by AI on operator direction.
Superseded — this item was already ratified and should not have been queued.
The earlier first-drain ratification request on this issue was raised in error. The admission gate dispatches a C3 bug-fix-shaped item when it carries either its own recorded ratification or a completed first drain — and this issue already carried recorded ratification in its body from attended triage on 2026-07-23. The drain cycle applied the blanket first-drain rule without first reading bodies for an existing ratification, so it queued an item that was already admitted.
Restored to the autonomous frontier:
agent-readyre-applied,needs-humanremoved. No human action is required, and this issue no longer appears in the attended queue's ratify view.kyle-sexton commentedon Jul 25, 2026
kyle-sexton commentedon Jul 25, 2026
This was generated by AI during work-loop execution.
Autonomous claim confirmed. This item was pre-gated by the work-loop's own admission pass before dispatch (
agent-ready+work-class: scoped, noneeds-human), so the interactive selection/confirmation prompt in/work-items:workwas satisfied by that prior admission decision rather than a user prompt — this comment records that per the skill's autonomous-invocation protocol. Re-verified immediately before claim: no open PR or branch referenced #792. Claimed via the seam (assignee + lease) and now implementing per the item's acceptance criteria in an out-of-tree worktree.feat(claude-ops): lane-launcher captures and persists the launch comm…