chore(pipeline-conductor): state the argv constraint in present tense

[dwu96]

Fixes #9372. Unblocks #9346.

Symptom

comment-history-baseline.json records 2 history-narration spans for
src/kiro_crew/builtin_skills/pipeline-conductor/scripts/fleet_probe.py, but the file on main
matches 3 of scripts/check_comment_history.py's patterns (L244 incident date 2026-08-30,
L255 "used to", L882 "no longer"). The gate is diff-scoped, so the gap surfaces as a
grew from 2 to 3 failure in Backend Lint & Type Check for whichever PR touches the file next —
live on #9346, whose failure log ends 0 new offender(s), 1 grown count(s), 0 file(s) with markers on added lines: proof that PR added none of the three.

Root cause

The checker and the baseline both arrived in a single commit (acfa55d0, #9328) whose baseline was
snapshotted against a tree older than the one it merged into; because the gate is deliberately
diff-scoped, #9328's own CI could not see drift in files #9328 did not touch. Nobody omitted
anything — a snapshot went stale in flight. (The issue as filed attributes the gap to aa75a403c /
#8736 leaving the entry "not raised"; that is chronologically impossible — #9328 landed ~1.5h
after aa75a403c — and this PR deliberately does not repeat it.)

fleet_probe.py is one corner of a 33-file drift; the full measured census is filed as #9384 and is
context, not scope.

Fix

One present-tense reword of the L882 docstring sentence:

• before: option/operand structure this function has to read is no longer recoverable.
• after: option/operand structure this function has to read cannot be recovered.

L882 is a false positive of the \bno longer\b pattern: the sentence states a property of the
NUL-to-space transformation (true of the current code, not narration about its past) — the same
distinction the checker's own docstring draws when it exempts a message reading "this token is no
longer valid" (rule-design follow-up: #9387). Rewording removes the match at zero cost to meaning.
The two genuine narration spans (L244, L255) stay: they are what the baseline entry of 2
legitimately carries.

Per the Main Ratchet Audit policy (#9350: "Fix the drift on main rather than raising the ceiling in
someone else's PR") and --write-baseline's own only-lowers contract, the entry is NOT raised, and
comment-history-baseline.json is untouched (no lowering needed — 2 is already the entry; and any
edit there would collide with #9364, which takes the same reword route for two other drifted files
and composes cleanly with this PR).

Verification

Red before green, with the gate's own entry points (not a re-implemented grep — the gate counts
distinct spans with overlaps collapsed):

• Red (file content at main 7177fa44, file in scope):
  ::error file=...fleet_probe.py::history narration in comments grew from 2 to 3 …
  comment-history gate FAILED: 0 new offender(s), 1 grown count(s), 0 file(s) with markers on added lines, 0 entr(y/ies) to lower.
  Corroborated by fix(pipeline-conductor): put the process age on the banned probe line #9346's Backend Lint & Type Check (3.12) CI log (same three lines, same trailer).
• Green (this branch):
  comment-history gate passed: nothing in scope narrates change history outside the baseline (7606 known marker(s) in 1720 file(s) still listed).
• Direct measurement (violations_in_source loaded from scripts/check_comment_history.py):
  3 spans before → 2 after, level with the baseline entry of 2.
• Mutation-check: no guard is added, so there is nothing to mutate. The equivalent revert-direction
  check was run: restoring the original sentence turns the gate red again (grew from 2 to 3),
  restoring the fix turns it green.
• Gates: check_black_formatting.py passed (scope: 1 changed file), flake8 clean, mypy clean on
  the changed file.
• Zero-regression: full backend suite (pytest -q -n auto --dist loadgroup, the identical bare
  invocation) on this branch and on a git worktree at origin/main; sorted FAILED/ERROR id sets
  compared both directions after ANSI stripping. Frontend cross-surface guard specs green.
  (Evidence detail below.)

No new unit test, deliberately: (a) the gate already enforces this exact invariant in CI for
every PR touching the file — that is precisely how the defect surfaced on #9346, so a unit test
would duplicate an existing enforced check; (b) the tempting stronger test — asserting tree-wide
that no baselined file exceeds its entry — would recreate the very anti-pattern the checker's
docstring rejects ("an unscoped gate reddens a PR because the base branch merged someone else's
file") and would redden every open PR today for all 33 drifted files (#9384). Closing the drift
hole properly is a design question, filed as #9385.

Follow-ups filed (scope drawn deliberately)

• comment-history baseline drift census: 33 files exceed or miss their entries on main #9384 — the measured 33-file / 55-span drift census (context for this fix, not its scope).
• comment-history gate: diff-scoping leaves main drift undetected until it reddens an innocent PR #9385 — the structural drift hole: how to detect main drift without reddening innocent PRs.
• comment-history gate: grown-count message accuses the contributor when the drift is inherited #9386 — the grown-count message reads as an accusation when the contributor's diff is clean.
• comment-history gate: "no longer" cannot tell descriptive docstring prose from history narration #9387 — descriptive vs historical "no longer": whether the checker's literal-string exemption
  reasoning should extend to descriptive docstring prose.

Sequencing

• fix(pipeline-conductor): put the process age on the banned probe line #9346 (the reddened PR) touches this file in the BANNED-probe print path; this PR touches one
  docstring sentence at L882 — no textual conflict expected. Landing this unblocks it.
• chore(comments): drop the history narration main outgrew its baseline on #9364 edits comment-history-baseline.json for chat_runner.py / context.py; this PR touches
  neither its files nor the baseline, so the two compose cleanly in either order.
• chore(dashboard): strip history narration from comments #9333 / chore(core): strip history narration from comments #9334 / chore(subsystems): strip history narration from comments #9338 do not touch fleet_probe.py (verified at branch time).

Pattern harvest

Rule candidate: when a ratchet gate and its baseline snapshot land as one commit, the snapshot must
be re-taken (or re-verified) against the merge target immediately before merge — a diff-scoped gate
cannot see its own baseline going stale in flight. Knowingly out-of-scope sibling sites: the other
32 drifted files measured in #9384, owned there rather than widened into this diff.

Screenshot evidence: none possible — a one-sentence docstring reword has no visual surface.

[github-actions]

Design Review (Fable 5) — ✅ PASS

Design-level review of 7a2f37fedbbb985de8b87203b885746860d0e183 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Minimal, zero-cost unblock of a stale-baseline gate failure, with the real root causes (drift hole, pattern over-breadth) correctly filed as follow-ups rather than smuggled in.

[DESIGN-REVIEWED] 7a2f37f

[github-actions]

Opus 4.8 Review — ✅ no blocking findings

Reviewed 7a2f37fedbbb985de8b87203b885746860d0e183 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 7a2f37f

Verdict parsed from the review's SHA-scoped output markers for commit 7a2f37fedbbb985de8b87203b885746860d0e183.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 7a2f37fedbbb985de8b87203b885746860d0e183: <one-sentence reason>

[github-actions]

First Principles Review (Fable 5) — ✅ PASS

Premise-level review of 7a2f37fedbbb985de8b87203b885746860d0e183 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

All verified: the \bno longer\b pattern exists at scripts/check_comment_history.py:98, the baseline entry for the file is 2, the file after the reword carries exactly the two declared spans (L244 date, L255 "used to"), the checker's own docstring draws the behavior-vs-narration distinction the author cites, and the new wording "cannot be recovered" matches none of the 18 patterns. The diff is exactly one line; nothing rides along. The 33-file sibling drift is declared, counted, and deferred to filed issues, with the fix's level stated explicitly — accepted-and-deferred under the contract.

First-Principles-Verdict: PASS

One point patch of a declared 33-file drift — confirm follow-up issues #9384/#9385 exist, since the general fix lives there, not here.

What this change ships

Inventory (1 item)

Intent: turn the comment-history gate green for fleet_probe.py so PR #9346 stops failing on drift it didn't cause — a FIX.

1. One docstring sentence reworded to present tense ("is no longer recoverable" → "cannot be recovered"), removing a gate match — justified

[FIRST-PRINCIPLES-REVIEWED] 7a2f37f

[github-actions]

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 7a2f37fedbbb985de8b87203b885746860d0e183 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 7a2f37f

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 7a2f37fedbbb985de8b87203b885746860d0e183: <one-sentence reason>

[iamwhatever]

Tier 1 auto-approve: chore (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: single-line docstring rewording in pipeline-conductor's fleet_probe argv-constraint note -- no executable statement changed, no runtime impact.

[chenmingwei23]

Tier 1 auto-approve: chore (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: chore(pipeline-conductor) one-line docstring wording change, no runtime impact.