fix(agents): shorten branch-plan-task description to invocation scope

[tvna]

Summary

Shortens agents/branch-plan-task.md's frontmatter description to
invocation-scope content only, fixing the 102-character YAML plain-scalar
truncation tracked in #1982.

Facts

• origin/main at 9838da14: raw description 717 chars, YAML-parsed 615
  chars -- reproduced directly (see fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982 for the exact command).
• The truncation is caused by a space followed by # inside the value
  (... and (Decision 20, issue #1476) ...), which YAML 1.1 reads as a
  plain-scalar comment marker.
• The rationale text this fix removes from the description
  (threat-model-and-authorization.md pointer + the Decision 20, issue #1476 citation) is not lost from the repository: it already appears in
  this same file's own body, line 53.
• hooks/gitapex_sync_opencode.py copies the description: line
  verbatim; out of scope per fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982's own Constraints, and unaffected
  since the new value round-trips cleanly for both runtimes.

Assumptions

• The issue author (repo OWNER), in a live session, chose deliberate
  shortening over the issue's two originally-drafted options (quote the
  value / rephrase the #1476 citation) -- recorded on fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982 itself
  under "Decided approach". This is treated as authoritative per
  planning-a-branch-from-an-issue's own Step 3 (an issue-owner comment
  narrowing scope wins over the original body).

Risk / blast radius

• User-visible: the shortened description is what every consumer of the
  gitapex:branch-plan-task / branch-plan-task agent-type listing sees
  from now on (Claude Code and, via the sync generator, OpenCode).
  Intentional, not incidental.
• No behavior change: nothing in agents/branch-plan-task.md's body,
  tool restrictions, or Sanctioned call sites list changes -- only the
  frontmatter summary text (plus documentation fixes from two Step 8
  review rounds, see Execution log).

Rollback

git revert the merge commit(s) -- single-line/small text changes, exact
prior text restored, no follow-up steps needed.

Verification

Acceptance Criteria Map (from #1982, "Decided approach" section):

Criterion	Interpretation	Planned ops	Proof method	Residual risk
description must contain only "when invoked" content	Keep "what it is + the two sanctioned call sites"; delete the rationale clause and the negative-constraint clause	Replace agents/branch-plan-task.md line 3 description value	YAML round-trip: parsed description == raw text, no # present -- PASS, confirmed against the merged commit (256/256 chars, no #)	#1982's unaudited skills/*/SKILL.md frontmatter risk is unchanged; hooks/gitapex_sync_opencode.py out of scope, unaffected

Commands run against the current head commit (72af1888):

$ python3 -c "import yaml; ..."   # round-trip check: raw==parsed==256, no '#' -- PASS
$ uv run --frozen python3 -m pytest tests/test_gitapex_sync_opencode.py -q
31 passed
$ uv run --frozen python3 -m ruff check tests/test_gitapex_sync_opencode.py agents/branch-plan-task.md docs/gitapex/plans/2026-09-13-claude-focused-fermat-c18ox6.md
All checks passed!
$ uv run --frozen python3 .github/scripts/gitapex_gate_local_preflight.py
local preflight: all 49 wired gate(s) passed.
$ uv run --frozen python3 .github/scripts/gitapex_gate_behind_base.py
OK: up to date with origin/main (4 commit(s) ahead).

Full repo suite also run in the dispatched task's own worktree before merge: uv run --frozen python3 -m pytest --no-cov -q -- 9161 passed. CI on the current head: all 34 checks green except independent-review-pending (this section is what satisfies it).

Checklist

• Tests pass locally
• Docs updated if behavior changed -- N/A, frontmatter-only change, no docs describe this description's exact wording
• Issue number cited in every commit
• If this PR adds/modifies a skills/*/SKILL.md, a docs/superpowers/specs/*.md design doc, a security-relevant skill, or a deterministic checker script (skills/*/scripts/*.py, evals/scripts/*.py, .github/scripts/*.py), a ## Skill audit evidence section discloses the required verdicts/waivers -- N/A, no such file touched
• If this PR adds a new Kept-edit-log entry to any evals/*/split.md, that entry discloses a Transfer check line -- N/A
• If this PR adds or increases a skills/*/SKILL.md's Stop-boundary bullets or named dispatch branches, evals/<skill>/tasks/*.yaml gained at least as many new fixtures -- N/A

Merge gate: independent review

This PR is also subject to the independent-review-pending required
status check (see .github/workflows/independent-review-pending.yml /
.github/scripts/gitapex_gate_independent_review_pending.py). It stays
pending/failing until a ## Independent review verdict section naming
this PR's current head commit is recorded in this body --
drafting-a-pr-to-merge's own Step 8 records it once its independent
review completes. There is nothing for you to do here now: do not
pre-fill this section yourself, and do not remove this note.

Independent review verdict

• Verdict: CLEAN
• Verified commit: 72af188

Outer layer (GitHub-native reviewer): requested a GitHub Copilot review
via request_copilot_review (no confirmation this repository has Copilot
review enabled -- no reviewer entry appeared in get_reviews afterward).
Disclosed per this skill's own asymmetric-guarantee note: Copilot's
review, if it runs, is Comment-only with no pass/fail signal of its own,
weaker than the "Claude Code Review" App's severity summary; this
repository's own installation of that App was not confirmed by the
operator, so it was not used. Not waited for; no response has arrived.

Inner layer: a fresh reviewing-an-artifact pass (effort: low), since
this PR's own Execution log carries no TaskCompleted entry matching
this exact head SHA with a run_id. Ran a 5-axis fan-out (correctness,
blast-radius, reuse-and-simplification, convention, security) against
the diff at head 7cd6436241ae936cee48c8cd5f6c4d4fd2f389ef:

• Correctness: 1 confirmed finding -- the plan doc's "No other line in
  the file changes" claim was false (a Step 8 follow-up fix also
  changed the file body). Fixed in 72af1888.
• Blast-radius: 2 findings. (a) Confirmed -- the plan doc's "already
  appears... (line 53)" claim overclaimed what was actually preserved
  and where. Fixed in 72af1888. (b) .claude/agents/branch-plan-task.md
  (a separate, intentionally-maintained project-local deployment
  variant) carries the identical #1476 truncation hazard. Not a
  defect in this diff -- pre-existing, out of issue fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982's own scope,
  and already disclosed in this PR's Residual risk section; a new,
  separate issue is being filed for it rather than folding it in here.
• Reuse-and-simplification: no confirmed findings (two low-confidence
  cosmetic notes, below the low-effort confidence bar).
• Convention: 1 confirmed finding -- the plan doc's filename did not
  follow the documented docs/gitapex/plans/<date>-<branch-name>.md
  shape. Fixed in 72af1888 (renamed to
  2026-09-13-claude-focused-fermat-c18ox6.md).
• Security: 0 confirmed findings. 1 disclosed unconfirmed-concern
  (security-tier, reported per this skill's own unconditional rule
  regardless of confidence): shortening the description reduces
  at-a-glance visibility into this deployment variant's weaker
  (prompt-only, no embedded hook) enforcement posture for a downstream
  reader who only skims the description: field. The full caveat
  remains intact in this same file's own body and cross-referenced
  correctly (see the correctness/blast-radius fixes above); this is a
  knowingly-accepted trade-off per the issue author's own recorded
  "Decided approach", not a new gap this diff introduces.

All three confirmed findings above were fixed and verified (ruff,
ASCII-only scan, plans-traceability CI check, and re-grep for stale
references) in commit 72af1888, the PR's current head. Step 1
re-classified the incremental fix diff (7cd64362 to 72af1888, prose
and a rename only, no code, no security-tier signal) as safe, per this
skill's own Step 1 rule -- no further fan-out was run against it.

Zero confirmed findings stand against the current head. The disclosed
unconfirmed-concern and the out-of-scope sibling-file observation are
recorded above rather than silently dropped, per this skill's own
never-silently-fold rule, but do not block this outcome.

Execution log

• PlanApproved -- Branch Plan and ACM for fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982 approved in-session by
  the issue author/repo OWNER (2026-09-13). Re-verification marker
  recorded on fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982.
• TaskStarted -- task-1-shorten-branch-plan-task-description dispatched
  to a worktree-isolated branch-plan-task subagent (2026-09-13).
• Screening: gitapex_check_canonical_governance_paths.py returned
  no-match (needs full-diff review); review-persona ran
  screening-a-low-trust-contribution checks 2-8 -- checks 2, 4-8 PASS;
  check 3 hard-flagged (diff modifies the governance/trust-anchor file
  agents/branch-plan-task.md) by that check's own unconditional design.
  Commit-message provenance scan: PASS (1 commit, clean).
• Human escalation per Step 7's failure-dispatch rule for a screening
  flag: presented the flag plus the fact the diff exactly matches the
  issue-author/repo-OWNER's own already-approved 256-char text; the
  owner confirmed proceeding with the merge in-session (2026-09-13).
• TaskCompleted -- task commit 7d10be52 cherry-picked onto the shared
  branch as 59fc3f14 and pushed. origin/main drift check: clean.
• Step 8 refactor/simplify pass (branch-plan-task, worktree-isolated):
  no behavior-preserving simplification needed; flagged one
  behavior-affecting finding out of its own scope (see next line).
• Step 8 independent adversarial review (review-persona): 2 findings.
  (1) agents/branch-plan-task.md line 74 pointed to "the reference
  cited above", but that citation was the exact clause this PR's
  description edit removed -- a dangling pointer. Fixed: now cites
  skills/executing-a-branch-plan/references/threat-model-and-authorization.md
  explicitly. (2) tests/test_gitapex_sync_opencode.py's docstring still
  described the truncation defect as present/tracked-separately; updated
  to past tense since this PR fixes it. Both fixed in commit 7cd64362,
  pushed; origin/main drift re-checked clean after the fix round.
  (3) A third, out-of-scope observation: .claude/agents/branch-plan-task.md
  (a separate, intentionally-maintained project-local deployment variant
  carrying its own embedded enforcement hooks) has the identical #1476
  truncation hazard in its own description. Confirmed with the repo
  owner in-session that the two files are deliberately different
  variants, not duplicates, and that this is real but out of fix(agents): branch-plan-task's description is truncated 102 characters by a YAML plain-scalar comment marker #1982's
  scope -- routed to a new, separate issue rather than folded into this
  PR.
• StageCompleted -- Step 8 clean (both confirmed findings fixed and
  verified), marking this PR ready for review.
• drafting-a-pr-to-merge Step 8: fresh reviewing-an-artifact pass
  (5-axis, low effort) against head 7cd64362 found 3 more confirmed
  findings and 1 disclosed unconfirmed-concern (see Independent review
  verdict above); all 3 confirmed findings fixed and pushed as
  72af1888. origin/main drift re-checked clean.

Related Issue

Closes #1982

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: d0d18f75-848a-43b5-8c06-0aa18e2f5cff

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[tvna]

The independent-review-pending check failing right now is expected, not a defect.

This PR is still under active execution by the executing-a-branch-plan skill (see the branch-plan-executing label) for #1982. Per the "Merge gate: independent review" note already in this PR's body, that check stays pending/failing until this skill's own Step 8 mandatory refactor/adversarial-review pass completes and records a ## Independent review verdict section here. No action needed on this check until then; it will resolve once Step 8/9 land.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.57%. Comparing base (9838da1) to head (72af188).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1995   +/-   ##
=======================================
  Coverage   99.57%   99.57%           
=======================================
  Files         171      171           
  Lines       29071    29071           
  Branches     3558     3558           
=======================================
  Hits        28948    28948           
  Misses        123      123           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.