fix(bin): prioritize active pipeline-owned crew runs

[kunchenguid]

Intent

PR0 of the captain-approved Firstmate startup/Bearings redesign: the F10 crew-state run-attribution precedence HOTFIX, which ships first ahead of the rest of the staged plan (implement ONLY PR0; PR1-PR4 and parallel fixes are explicitly out of scope, and the home-summary invalidity model must NOT change in this PR - that is later follow-up; the hotfix alone stops the cascade's trigger). Problem (confirmed live): bin/fm-crew-state.sh binds a superseded FAILED no-mistakes run to a task instead of the LIVE replacement run, because the live run's pipeline-owned lane head is not a git object in the task worktree: fm_nm_head_matches_worktree rejects it, then the coarse fallback nm_runs_status_for_branch continues past the RUNNING row (head unresolvable) onto the older FAILED row whose head equals the stale worktree HEAD. Observed on task fm-actionable-update-escalation-misclassification-r1: failed run at d0bde839 chosen over the running run at a pipeline-owned head; the home summary then flips valid=false (terminal_in_flight) and Bearings hides the home's live work. Required precedence: (1) a daemon-attributed ACTIVE run for the task's branch (from axi status in the worktree, matching branch name) is AUTHORITATIVE - head equality is NOT required while branch_sync.state is pipeline_owned; report working with the run-step detail. (2) Otherwise the newest terminal run for the branch that no later run supersedes surfaces its terminal state; a genuinely-failed run with NO later run on the branch MUST still report failed - hiding real failures is equally wrong. (3) An unresolvable run head classifies as unknown attribution and must NEVER cause a silent fallback onto an older row; in the coarse no-mistakes runs scan, an ACTIVE row for this branch whose head cannot be resolved must not be skipped in favor of an older terminal row. (4) Scope is bin/fm-crew-state.sh plus bin/fm-nm-run-lib.sh only for genuinely needed helpers. Implementation decisions made deliberately: the new attribution helpers (fm_nm_head_resolvable, fm_nm_branch_sync_state, fm_nm_run_is_active, fm_nm_run_is_pipeline_owned_active) live in bin/fm-nm-run-lib.sh because that lib is the declared single owner of the run-attribution rule (one-owner rule from firstmate-coding-guidelines); the pipeline_owned exemption deliberately requires an ACTIVE run (a terminal run has released the branch, and binding one by branch name alone would recreate the historical reused-branch misattribution the head rule prevents); the coarse-scan guard stops on ANY unresolvable head for the task's branch (unknown attribution) while a resolvable-but-mismatched head keeps the existing historical reused-branch skip semantics; fm-teardown.sh's stricter head-equality use of the lib is deliberately unchanged in this PR. Acceptance tests (T1, both directions, required): fixture worktree at head H1 on branch B with a PATH-stubbed no-mistakes whose axi status answers a RUNNING run on B at unresolvable head H2 with branch_sync.state pipeline_owned and whose runs list shows newest running-B-H2 and older failed-B-H1 must report working with source run-step; the same fixture where the stub reports ONLY failed B H1 (no later run) must still report failed; existing crew-state tests stay green; tests exercise observable behavior through the real fm-crew-state.sh entrypoint with a PATH-stubbed no-mistakes, never implementation-string matching. The test fixture's branch_sync TOON shape was copied from the live incident run's real axi status output, and every fixture head is deliberately unresolvable so only the pipeline-owned exemption can attribute (guarding against accidental field grabs). Two extra negative-control tests pin that the exemption requires branch_sync.state=pipeline_owned specifically and never applies to a terminal run. All tests were verified to fail before the fix and pass after it.

What Changed

• Attribute active pipeline_owned no-mistakes runs by branch without requiring their lane head to resolve in the task worktree.
• Stop coarse run scans at an unresolvable newer head instead of falling through to a superseded terminal run, while preserving genuine terminal failures.
• Add entrypoint regression coverage and update run-attribution documentation.

Risk Assessment

✅ Low: The change is narrowly scoped, preserves terminal-run attribution, and adds behavioral regression coverage for the pipeline-owned precedence and coarse-scan guard.

Testing

After inspecting the clean target diff, I ran the focused crew-state behavior suite, reproduced the original base-commit misattribution, and exercised the real CLI entrypoint end-to-end: an active pipeline-owned run now reports working despite an unresolvable head and older failed row, while a genuine unsuperseded failure still reports failed.

Evidence: F10 fixed behavior: active pipeline-owned run wins, genuine failure remains visible

Source: F10 fixed behavior: active pipeline-owned run wins, genuine failure remains visible

F10 end-to-end evidence (real bin/fm-crew-state.sh entrypoint, PATH-stubbed no-mistakes)
Scenario 1: newest ACTIVE pipeline-owned run has an unresolvable lane head; older row is failed at the worktree head.
state: working · source: run-step · validating (running)
Scenario 2: only the genuine failed run remains on the branch at the worktree head.
state: failed · source: run-step · run failed

Evidence: Base-commit regression reproduction: older failed run incorrectly surfaced

Source: Base-commit regression reproduction: older failed run incorrectly surfaced

Regression reproduction against base commit 4f89f5b (same real entrypoint logic, PATH-stubbed no-mistakes):
Newest row is ACTIVE pipeline-owned at an unresolvable head; older row is failed at the worktree head.
state: failed · source: run-step · run failed

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

• Inspected the target diff and confirmed the worktree remained clean with git status --short.
• Ran the focused behavior suite: tests/fm-crew-state.test.sh.
• Executed bin/fm-crew-state.sh end-to-end with a PATH-stubbed no-mistakes for both required precedence directions and captured the CLI transcript.
• Executed the base commit's crew-state logic against the live-run/older-failure fixture to reproduce the original false failed result.
• Checked changed files for patch whitespace errors with git diff --check 4f89f5b5e235469d32c037b7792d6dba5bdc272d..002b0f98cd8e447f988059750b134d8e6f6cb000 -- bin/fm-crew-state.sh bin/fm-nm-run-lib.sh tests/fm-crew-state.test.sh.

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

Confidence Score: 5/5

The PR appears safe to merge with no actionable defects identified.

The changed attribution path is narrowly gated to active pipeline-owned same-branch runs, preserves strict terminal attribution, and has behavioral coverage for both precedence directions and negative controls.

_{Reviews (1): Last reviewed commit: "no-mistakes(document): Updated run-attri..." | Re-trigger Greptile}