[Phase 3] Extend Fleet Monitor stub-drift coverage to the initiative-driver stub

[github-actions]

Story

As a fleet maintainer,
I want have Fleet Monitor track per-repo initiative-driver stub coverage/drift the same way it already tracks the planner stub,
so that an out-of-date or missing driver stub in an enrolled repo surfaces in the daily health signal instead of failing silently.

Acceptance Criteria

1. scripts/fleet_monitor.sh detects drift for the driver stub .github/workflows/initiative-driver.yml against petry-projects/.github/standards/workflows/initiative-driver.yml, in ADDITION to the existing planner stub — generalized to iterate a list of (name, stub_path, canonical_path) entries rather than a single hardcoded STUB_PATH/CANONICAL_STUB_PATH.
2. The fleet report renders a driver stub coverage/drift section (reusing fleet_stub_drift.sh's pure helpers), and drifted driver stubs feed the alert JSON / HAS_STUB_DRIFT path.
3. The actions-fleet-monitor.yml stub-drift alert step covers the driver stub without regressing the planner-stub coverage (separate section/issue per stub kind, or a clearly-labelled combined section).
4. Bats cover the multi-stub generalization (extend tests around fleet_stub_drift / fleet_monitor) and confirm the existing planner-stub behaviour is unchanged.
5. If the canonical driver-stub SHA cannot be read (the template is not yet merged in petry-projects/.github), the driver registry entry is skipped with a ::warning:: (per-stub, non-fatal) — exactly as the existing single-stub path does when its canonical SHA is empty — so Fleet Monitor degrades gracefully before and after the cross-repo template lands.

Tasks / Subtasks

• Generalize fleet_monitor.sh: replace the single STUB_PATH/CANONICAL_STUB_PATH (+ the per-repo loop ~:187-206 and report/alert sections) with iteration over a stub registry including both initiative-planner and initiative-driver. (AC: test issue from agent #1, Go-live improvements for PR review agent #2)
• Keep fleet_stub_drift.sh's pure helpers (classify_stub_drift / stub_drift_row / stub_drift_alert_json / generate_stub_drift_report) reusable per stub; parameterize the report heading by stub name. (AC: Go-live improvements for PR review agent #2)
• Update actions-fleet-monitor.yml's stub-drift alert step to surface driver-stub drift additively (do not drop the planner section/issue). (AC: Add @claude delegation, auto-merge, and rebase handling #3)
• Add/extend bats for the multi-stub path and a regression assertion that planner-stub coverage is unchanged. (AC: Optimize review: small-PR and incremental fast paths #4)
• Preserve the per-stub warn-and-skip when a canonical SHA is empty (fleet_monitor.sh ~:198 already does this for one stub) so a not-yet-merged driver template degrades gracefully rather than failing the run. (AC: feat: add Copilot engine support via REVIEW_ENGINE toggle #5)

Dev Notes

• fleet_stub_drift.sh is ALREADY pure and stub-agnostic (takes a canonical SHA + a per-repo SHA, classifies ALIGNED/DRIFTED/MISSING). The hardcoding lives in fleet_monitor.sh: STUB_PATH (:35), CANONICAL_STUB_PATH (:37), the comparison loop (:187-206), the report section (:239-243), and the alert export (~:309-319).
• Generalize by looping over a small registry, e.g. [(initiative-planner, .github/workflows/initiative-planner.yml, standards/workflows/initiative-planner.yml), (initiative-driver, .github/workflows/initiative-driver.yml, standards/workflows/initiative-driver.yml)], calling the existing pure helpers per entry and accumulating per-stub drift files.
• Keep it ADDITIVE: the planner stub section, its tracking issue, and HAS_STUB_DRIFT semantics must keep working — the Live idea:approved canary + Fleet Monitor coverage #822 behaviour is load-bearing.
• The driver canonical path must match Story 2's exactly; this story is blocked_by Story 2 so the canonical template exists to compare against (otherwise every enrolled repo reads as MISSING/404).
• Cross-repo merge ordering: this story needs Story 2's stub template MERGED to petry-projects/.github's default branch (not just authored) for a non-empty canonical SHA — blocked_by [2] captures the intra-plan edge, but the canonical read is cross-repo, so the generalized loop must keep the existing empty-canonical warn-and-skip per stub.

Project Structure Notes

All in .github-private — dev-lead-authorable. Net-new bats should be added to lint.yml's bats list per AGENTS.md.

References

• scripts/fleet_monitor.sh#stub-coverage-and-drift
• scripts/fleet_stub_drift.sh
• .github/workflows/actions-fleet-monitor.yml#track-stub-drift
• tests/ (fleet monitor / stub drift bats)

Likely target surface

• scripts/fleet_monitor.sh
• scripts/fleet_stub_drift.sh
• .github/workflows/actions-fleet-monitor.yml
• tests/

Story prepared by the BMAD Scrum Master (Bob) for epic #882. Status: ready-for-dev.

[don-petry]

Dev-Lead Implementation Plan

Issue: #886 — [Phase 3] Extend Fleet Monitor stub-drift coverage to the initiative-driver stub

Scope

Generalize Fleet Monitor's single hardcoded planner-stub drift check into iteration over a small stub registry (initiative-planner + initiative-driver), reusing fleet_stub_drift.sh's pure helpers per stub, with a per-stub warn-and-skip when a canonical SHA is unreadable.

Implementation Checklist

• scripts/fleet_stub_drift.sh: parameterize generate_stub_drift_report by a stub display-label (default keeps "Initiative-planner"); extend stub_drift_alert_json to optionally tag each row with stub/stub_file fields.
• scripts/fleet_monitor.sh: replace STUB_PATH/CANONICAL_STUB_PATH + the single per-repo loop/report/alert export with a STUB_REGISTRY of (name, label, stub_path, canonical_path); iterate it, building per-stub drift files, a report section per stub, and a merged alert JSON; preserve the empty-canonical ::warning:: warn-and-skip per stub.
• .github/workflows/actions-fleet-monitor.yml: update the "Track stub drift" step to group the merged alert JSON by stub and create/update a separate tracking issue per stub kind (planner issue title unchanged); keep HAS_STUB_DRIFT/STUB_DRIFT_COUNT semantics.

Tests to Write

• tests/fleet_stub_drift.bats: generate_stub_drift_report renders a passed stub label (e.g. "Initiative-driver"); default label still renders "Initiative-planner" (regression).
• tests/fleet_stub_drift.bats: stub_drift_alert_json with a stub label tags each DRIFTED row with stub/stub_file; without a label the existing output is unchanged (regression).

Test Command

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

[don-petry]

Dev-Lead: Implementation Complete

Plan Execution

• Parameterized fleet_stub_drift.sh pure helpers: generate_stub_drift_report now takes a stub-label (default Initiative-planner); stub_drift_alert_json optionally tags each DRIFTED row with stub/stub_file.
• Generalized fleet_monitor.sh: replaced the single STUB_PATH/CANONICAL_STUB_PATH with a STUB_REGISTRY of (name, label, stub_path, canonical_path) covering both initiative-planner and initiative-driver; iterates it, builds a per-stub drift file, a report section per stub, and a merged alert JSON. Per-stub empty-canonical ::warning:: warn-and-skip preserved (AC#5).
• Updated actions-fleet-monitor.yml "Track stub drift" step to group the merged JSON by stub kind and create/update a separate issue per stub (planner issue title unchanged — additive, AC#3).
• Extended tests/fleet_stub_drift.bats for the parameterized helpers + a regression test that the default heading is still Initiative-planner.

Test Results

ok 9 stub_drift_alert_json: emits only DRIFTED enrolled repos
ok 10 stub_drift_alert_json: an all-aligned TSV yields an empty array
ok 11 stub_drift_alert_json: an empty/absent TSV yields an empty array
ok 12 stub_drift_alert_json: a stub label/file tags each DRIFTED row (multi-stub)
ok 13 generate_stub_drift_report: summarizes counts and lists drifted repos
ok 14 generate_stub_drift_report: a passed stub label headlines the section (multi-stub)
ok 15 generate_stub_drift_report: the default label is still Initiative-planner (regression)
ok 16 generate_stub_drift_report: a clean fleet reports no drift

Full lint.yml bats suite (686 tests) passes; bash .dev-lead/scripts/dev-lead-lint.sh passes (shellcheck + agent-profile validation).

Files Changed

• scripts/fleet_stub_drift.sh: parameterized report heading + alert-JSON stub tagging (pure, backward-compatible defaults).
• scripts/fleet_monitor.sh: single-stub path → stub registry iteration with per-stub drift files, report sections, merged alert JSON, and per-stub warn-and-skip.
• .github/workflows/actions-fleet-monitor.yml: stub-drift alert step now tracks each stub kind in its own issue; clean-run notice mentions both stubs.
• tests/fleet_stub_drift.bats: +3 tests (driver label, multi-stub alert tagging, planner-default regression).

Notes

• Tests live in the already-listed tests/fleet_stub_drift.bats, so no lint.yml bats-list change was needed.
• Graceful degradation (AC#5): if the canonical driver template is not yet merged in petry-projects/.github, that registry entry is skipped with a per-stub ::warning:: and the planner path runs unchanged; if both are unreadable, the monitor emits [] and annotates a clean run.
• No new AGENTS.md workflow exception needed — this modifies the existing canonical fleet-monitor workflow rather than adding a new one.