[Fleet Monitor] petry-projects/TalkTerm — .github/workflows/dev-lead.yml

[don-petry]

🟠 DEGRADED: petry-projects/TalkTerm / .github/workflows/dev-lead.yml

Metric	Value
Failure Rate	33.3% (2 / 6 runs)
Status	DEGRADED
p50 Duration	3s
p95 Duration	15s
Successful Runs	2
Cancelled Runs	0

Workflow: .github/workflows/dev-lead.yml

Last updated by Fleet Monitor run on 2026-07-06.
Threshold: failure rate > 10% over the monitored window.

[don-petry]

Dev-Lead Implementation Plan

Issue: #275 — [Fleet Monitor] petry-projects/TalkTerm — .github/workflows/dev-lead.yml

Root Cause

The degraded failure rate traces to .github/workflows/dev-lead.yml being a stale, non-compliant thin stub:

• It pins the reusable workflow to a frozen SHA 5a9a247 (comment claims # main) that is now 66 commits behind main and 63 behind the promoted dev-lead/stable channel — missing engine-reliability fixes (gemini quota/stderr handling Compliance: check-suite-auto-trigger-1236702 #226, heredoc INTENT_BODY #559, model-selection #556/#569, hands-off label, worktree release #470, etc.).
• It is missing the fork-PR skip guard, so PRs from forks (no secrets) fail the required check.
• It is missing agent_ref: dev-lead/stable and pins to a SHA rather than the @dev-lead/stable channel — exactly the dev-lead-stub-pin / dev-lead-stub-agent-ref compliance violations that show up as failing labeled-issue runs.

The canonical "copy verbatim" stub (.github-private/.github/workflows/dev-lead.yml@main) has since moved to the self-host @dev-lead/stable channel pin (#535/#506), which decouples the agent from broken-main and brings in the accumulated reliability fixes.

Scope

Re-align .github/workflows/dev-lead.yml to the canonical upstream thin stub (verbatim), adopting the @dev-lead/stable channel pin, agent_ref, and the fork-PR skip guard.

Implementation Checklist

• Replace .github/workflows/dev-lead.yml with the canonical verbatim stub from petry-projects/.github-private/.github/workflows/dev-lead.yml@main
  • Pin uses: → dev-lead-reusable.yml@dev-lead/stable (drop stale SHA)
  • Add with: agent_ref: dev-lead/stable
  • Add fork-PR if: skip guard
  • Sync repository_dispatch types (dev-lead-reviews-retry)
  • Drop stale concurrency comment block + branches: [main] filter to match canonical

Tests to Write

• No unit-test framework covers this CI YAML. Per Phase 2 exception this is a behavior-preserving CI configuration change; verification is via:
  • python3 -c yaml.safe_load parse check of the resulting workflow
  • Byte-equality diff against the canonical upstream stub (must be empty)
  • .dev-lead/scripts/dev-lead-lint.sh

Test Command

bash .dev-lead/scripts/dev-lead-lint.sh (+ full npm test for regression)

[don-petry]

Dev-Lead: Implementation Complete

Plan Execution

• Identified root cause: .github/workflows/dev-lead.yml was a stale, non-compliant thin stub
• Replaced it with the canonical verbatim stub from petry-projects/.github-private/.github/workflows/dev-lead.yml@main
  • Pin moved off stale SHA 5a9a247 (66 commits behind main / 63 behind stable) → channel ref dev-lead-reusable.yml@dev-lead/stable
  • Added with: agent_ref: dev-lead/stable
  • Added fork-PR if: skip guard
  • Synced repository_dispatch types (added dev-lead-reviews-retry)
  • Dropped stale concurrency comment block and branches: [main] filter to match canonical
• Verified resulting file is byte-identical to canonical upstream stub
• Verified the @dev-lead/stable reusable workflow declares every input/secret/permission the stub passes (agent_ref confirmed present)

Why this fixes the degraded failure rate

The frozen SHA pin missed ~63–66 commits of engine-reliability fixes (gemini quota/stderr handling #226, heredoc INTENT_BODY #559, multi-tier model selection #556/#569, stale worktree release #470, dev-lead:hands-off exclusion). It also lacked the fork-PR skip guard (fork PRs have no secrets → required check fails) and violated the dev-lead-stub-pin / dev-lead-stub-agent-ref compliance rules, which surfaced as failing labeled-issue runs. The @dev-lead/stable self-host channel (#535/#506) additionally decouples the agent from broken-main, so a bad change on main can no longer gate its own fix.

Test Results

$ python3 -c "import yaml; yaml.safe_load(open('.github/workflows/dev-lead.yml'))"
YAML OK

$ diff .github/workflows/dev-lead.yml <canonical upstream stub>
IDENTICAL

$ bash .dev-lead/scripts/dev-lead-lint.sh
  [lint] shellcheck on 1 file(s)...
  [lint] all checks passed
EXIT=0

Files Changed

• .github/workflows/dev-lead.yml: re-aligned thin caller stub to canonical upstream — @dev-lead/stable channel pin + agent_ref, fork-PR skip guard, synced dispatch types (1 file, +15 / -9).

Notes

• TDD exception (Phase 2): behavior-preserving CI-configuration change; no application code touched and no unit-test harness covers workflow YAML. Verification was YAML-parse + byte-equality against the authoritative upstream stub + dev-lead-lint.
• The repo JS/Vitest suite is unaffected by this CI-YAML change (no src/ files touched); node_modules is not installed in this environment, so the Electron suite was not exercised.
• No unrelated files modified.