[kyle-sexton]

Refs: #6611

Summary

Restores the 650-character naming:name-it-better description that #12186 shortened to 624 characters. A retroactive negative control showed the eval that approved the cut cannot tell the shipped text from a one-line generic description. So the cut was never measured, and under the owner's rule it does not stay.

Fix

• plugins/naming/skills/name-it-better/SKILL.md: line 2 returns to its text before feat(naming): shorten the name-it-better description #12186. Nothing else in the file changes.
• A patch changelog fragment for the naming plugin.

Verification

The control ran on Sonnet 5.5, Claude Code 2.1.316. The shipped description and description: "Helps with naming." were launched together, and all 12 passes were VALID.

Suite	shipped	broken
Routing, 4 cases × 5 runs (merged mean)	0.967	0.983
Held-out trigger: should-trigger fired	24/24	24/24
Held-out trigger: should-not quiet	22/24	23/24

• Routing: the noise report over the 4 merged cases called the broken copy non-inferior, with no case drop.
• Trigger: one case drop, on a single false fire in 3 runs, did not reproduce in a preregistered confirmation run.
• git diff origin/main -- plugins/naming/skills shows only the description line.

Related

• feat(naming): shorten the name-it-better description #12186: the cut this reverts.
• Concise skills rollout: naming (1 skills, 1 model-invoked) #6611: control result. The issue stays open for a competitor-loaded routing suite.
• docs(playbooks): require a negative control before a rewrite's measurement counts #12222: the negative-control step in rewrite-a-plugin.md.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-02T03:09:36.353026Z	559d884	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

claude-review reviewed 559d8848fd8f383e87998407caa7c4ef299fa33e: read 2 of 2 changed file(s); 0 finding(s), 0 as inline comments.

Code review: no findings

This PR undoes #12186's shorter name-it-better description and adds a patch changelog fragment. I read every hunk of both changed files.

• plugins/naming/skills/name-it-better/SKILL.md: only line 2 (description) changes. I counted the restored text by hand and it is exactly 650 characters, so the PR and the fragment have that number right. All frontmatter fields stay the same, including argument-hint, user-invocable and disable-model-invocation. The restored text still says what the skill does, when to use it (an undecided name or a retry on a rejected one), the "never auto-locks a name" rule, the redirect to rename-references, and the tournament argument.
• .changes/naming/revert-6611-name-it-better-description-b4e2c7a1.md: it is laid out like the other fragments under .changes/: front matter with bump:, then a ### Changed bullet. A wording-only change to a description fits patch.
• Path-scoped rules: .claude/rules/skill-bodies-state-current-rules.md and .claude/rules/cost-claims.md match the SKILL.md path. The restored description restates no upstream specifics and makes no cost claims, so neither rule is broken.
• No closing issue: the PR says Refs: #6611, which does not close it, so there are no acceptance criteria to check.
• No plugin.json or marketplace.json changes, so the plugin-manifest checks don't apply.
• This lane has no earlier findings on this PR (round 0 of 3).

The routing and trigger eval numbers in the PR description come from the author. I did not run them.

Spend: none. This change uses no credits and creates no charges.

[github-actions]

claude-security-review reviewed 559d8848fd8f383e87998407caa7c4ef299fa33e: read 2 of 2 changed file(s); 0 finding(s), 0 as inline comments.

Security review: PR #12235

I found no security issues. I read every hunk of both changed files.

• plugins/naming/skills/name-it-better/SKILL.md: the only change is the description: frontmatter on line 2, which goes back to its earlier text. I checked it as an instruction-surface edit by comparing the removed and added lines. Both still say the skill never auto-locks a name and that a decided rename routes to the rename-references sweep, so no rule is removed or softened. The tool grants, user-invocable, disable-model-invocation and the argument hint are all unchanged.
• .changes/naming/revert-6611-name-it-better-description-b4e2c7a1.md: a new changelog fragment containing only prose. Nothing runs from it.

Neither file matches a trusted-path glob. No plugin manifest, workflow, trust control, token scope or merge hold is touched. The PR description has no embedded directives aimed at this lane.

critical-findings: 0