Live idea:approved canary + Fleet Monitor coverage

[don-petry]

Story

As a org owner,
I want add a post-merge canary on the idea:approved → dispatch path and extend Fleet Monitor coverage to the new stub,
so that a silent revert like #655 is caught on the first approval rather than when a human notices nothing planned.

Acceptance Criteria

1. A scheduled smoke check asserts the stub dispatches the central planner (a dry-run dispatch against a fixture Discussion, or a synthetic idea:approved add/remove) and alerts on a failure or the 'Unsupported event type: discussion' error.
2. The Fleet Monitor health-check ([Fleet Monitor] petry-projects/.github-private — .github/workflows/initiative-planner.yml #627) is extended to the new standards/workflows/initiative-planner.yml stub across enrolled repos.
3. An alert fires if any enrolled repo's stub drifts from the canonical template SHA.

Tasks / Subtasks

• Add a scheduled canary workflow/job that exercises the idea:approved → dispatch path in dry-run and alerts on failure. (AC: test issue from agent #1)
• Extend the Fleet Monitor health-check to track the new stub. (AC: Go-live improvements for PR review agent #2)
• Add stub-SHA drift detection to the monitor. (AC: Add @claude delegation, auto-merge, and rebase handling #3)

Dev Notes

• Directly closes bug: stale-base merge (#606) silently reverted the initiative-planner discussion→dispatch fix (#618); idea:approved stopped planning #655 prevention item 3 ('smoke-monitor the live trigger') — the discussion[labeled] path has no post-merge canary today.
• [Fleet Monitor] petry-projects/.github-private — .github/workflows/initiative-planner.yml #627 is the existing [Fleet Monitor] health-check tracking initiative-planner.yml; extend it rather than building a parallel monitor.
• Prefer a dry-run dispatch over a real plan so the canary creates nothing.

Project Structure Notes

Monitoring lives in .github-private alongside the planner; reuse the existing fleet-monitor/health-check tooling.

References

• [Fleet Monitor] petry-projects/.github-private — .github/workflows/initiative-planner.yml #627
• bug: stale-base merge (#606) silently reverted the initiative-planner discussion→dispatch fix (#618); idea:approved stopped planning #655

Likely target surface

• .github/workflows/initiative-planner.yml
• scripts/initiative-planner/

Story prepared by the BMAD Scrum Master (Bob) for epic #817. Status: ready-for-dev.

[github-actions]

Thanks for the detailed story and acceptance criteria — this is well-scoped and easy to pick up! Before we start implementation: for the scheduled canary (AC #1), should the dry-run dispatch target a fixture Discussion in this repo, or would you prefer a fully synthetic event (e.g., mocking the discussion[labeled] webhook payload) so no real Discussion object is needed at all?

[don-petry]

Dev-Lead Implementation Plan

Issue: #822 — Live idea:approved canary + Fleet Monitor coverage

Scope

Add a scheduled dry-run canary for the idea:approved → planner-dispatch path, and extend the Actions Fleet Monitor to track the per-repo initiative-planner.yml stub and alert on drift from the canonical template SHA.

Implementation Checklist

• AC#1 scripts/initiative_canary.sh — dry-run dispatch of initiative-planner.yml against a fixture Discussion, poll the resulting run, classify success / failure / Unsupported event type: discussion. Pure helpers + main() (mirrors auto_rebase_health.sh).
• AC#1 .github/workflows/initiative-planner-canary.yml — scheduled (daily) + workflow_dispatch canary; alerts (find-or-create issue) on a non-OK result.
• AC#2/Add @claude delegation, auto-merge, and rebase handling #3 scripts/fleet_stub_drift.sh — sourced lib of pure functions: classify a repo stub blob SHA vs. the canonical template SHA (ALIGNED / DRIFTED / MISSING), build report + alert JSON.
• AC#2/Add @claude delegation, auto-merge, and rebase handling #3 Extend scripts/fleet_monitor.sh — new "Initiative-planner stub coverage & drift" section reusing the discovered repo list; fetch canonical + per-repo stub SHAs, emit fleet_stub_drift.json, set HAS_STUB_DRIFT.
• AC#3 Extend .github/workflows/actions-fleet-monitor.yml — alert step gated on env.HAS_STUB_DRIFT == 'true'.
• Register the two new test files in .github/workflows/lint.yml (per AGENTS.md bats-list rule).

Tests to Write

• tests/test_initiative_canary.bats — classify_canary_run (OK / FAILED / UNSUPPORTED_EVENT / NO_RUN), alert-decision helper, alert-body render, and a gh-mocked main asserting the dry-run dispatch (dry_run=true).
• tests/fleet_stub_drift.bats — classify_stub_drift, row builder, report generator, alert-JSON, and count helpers over a fixture TSV.

Test Command

scripts/run-bats.sh tests/test_initiative_canary.bats tests/fleet_stub_drift.bats (full suite via lint.yml bats job); lint via bash .dev-lead/scripts/dev-lead-lint.sh

[don-petry]

Dev-Lead: Implementation Complete

Plan Execution

• AC#1 — scripts/initiative_canary.sh + .github/workflows/initiative-planner-canary.yml: daily (06:30 UTC) + workflow_dispatch canary that fires a dry-run dispatch of initiative-planner.yml against a fixture Discussion, polls the resulting run, and alerts (find-or-create tracking issue + wakes dev-lead) on a failure or the Unsupported event type: discussion regression fingerprint. Dry-run ⇒ creates nothing.
• AC#2 — Extended the Actions Fleet Monitor (scripts/fleet_monitor.sh §2a) to track each enrolled repo's initiative-planner.yml stub, reusing the already-discovered repo list (extends [Fleet Monitor] petry-projects/.github-private — .github/workflows/initiative-planner.yml #627 rather than building a parallel monitor). Pure classification logic in the new sourced lib scripts/fleet_stub_drift.sh.
• AC#3 — Drift detection: per-repo stub blob SHA compared against the canonical org-template SHA (ALIGNED/DRIFTED/MISSING). Drifted enrolled repos are exported to fleet_stub_drift.json; actions-fleet-monitor.yml opens/updates a fleet-stub-drift tracking issue and fires repository_dispatch to wake dev-lead when HAS_STUB_DRIFT.
• Registered both new test files in lint.yml; added an AGENTS.md "Workflow Files" exception entry for the repo-specific canary workflow.

Test Results

# New tests (TDD — written before implementation):
tests/test_initiative_canary.bats ... 14 ok
tests/fleet_stub_drift.bats ........ 13 ok

# Full lint.yml bats suite:
631 tests, 0 failures

# Lint:
bash .dev-lead/scripts/dev-lead-lint.sh  → all checks passed (shellcheck 69 files + agent profiles)
shellcheck --severity=warning -x scripts/{fleet_monitor,fleet_stub_drift,initiative_canary}.sh → clean
YAML parse (initiative-planner-canary.yml, actions-fleet-monitor.yml, lint.yml) → OK

Files Changed

• scripts/initiative_canary.sh (new): canary — pure classify_canary_run/canary_is_failure/canary_report + main() (dispatch+poll). Mirrors auto_rebase_health.sh layout.
• .github/workflows/initiative-planner-canary.yml (new): scheduled canary workflow with failure alerting.
• scripts/fleet_stub_drift.sh (new): pure stub-coverage/drift lib (classify, row, count, alert-JSON, report).
• scripts/fleet_monitor.sh: §2a stub-SHA scan over discovered repos; appends a "stub coverage & drift" report section; exports fleet_stub_drift.json + HAS_STUB_DRIFT/STUB_DRIFT_COUNT.
• .github/workflows/actions-fleet-monitor.yml: "Track initiative-planner stub drift" + "Dispatch dev-lead for stub drift" steps; tightened the clean-run annotation.
• .github/workflows/lint.yml: registered the two new bats files.
• tests/test_initiative_canary.bats, tests/fleet_stub_drift.bats (new): unit tests (gh mocked; no network).
• AGENTS.md: documented the repo-specific canary workflow exception.

Notes

• Per Live idea:approved canary + Fleet Monitor coverage #822 Dev Notes, the canary uses a dry-run dispatch (dry_run=true) so apply-plan only logs intended mutations to an artifact — no epic/issues are created.
• The canary requires GH_PAT_WORKFLOWS (a workflow_dispatch fired with GITHUB_TOKEN is accepted but never starts a run) and reads the fixture Discussion from repo variable INITIATIVE_CANARY_DISCUSSION; an unset fixture skips with a ::warning:: rather than failing.
• "Enrolled" = a repo that has the .github/workflows/initiative-planner.yml stub. MISSING (no stub) is reported as informational/not-enrolled and is not alerted; only DRIFTED enrolled repos trigger the drift alert (AC#3).
• Stub-drift compares GitHub contents-API blob SHAs (git blob object IDs), so equal SHAs ⇒ byte-identical stubs. Canonical source defaults to petry-projects/.github standards/workflows/initiative-planner.yml (overridable via CANONICAL_STUB_REPO/CANONICAL_STUB_PATH).
• Per repo constraints, no commits/PRs were created — the CI workflow handles git operations.