fix(skills): expand skill-creator description to cover edit/audit/review triggers by haynzz · Pull Request #39158 · openclaw/openclaw

haynzz · 2026-03-07T20:10:01Z

Summary

Problem: The skill-creator skill description only listed creation-oriented phrases ("Create or update AgentSkills…designing, structuring, or packaging"). When a user said "tidy up a skill", "audit the skill", "review the skill", "clean up the skill", or "improve this SKILL.md", the model did not recognise the skill as relevant and winged the task without the skill's guidance.
Why it matters: Skill maintenance (editing, auditing, restructuring existing skills) is at least as common as creating new ones. Skipping the skill for these cases causes inconsistent results and undermines the skill's value as a spec-enforcing guide.
What changed: The description field in skills/skill-creator/SKILL.md frontmatter is expanded to explicitly cover edit/improve/review/audit/tidy-up scenarios and lists the exact trigger phrases.
What did NOT change: No logic, scripts, body copy, or other files were modified. This is a single-line frontmatter edit.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Closes skill-creator: description too narrow — fails to trigger on edit/refactor tasks #36039

User-visible / Behavior Changes

The skill-creator skill now triggers for maintenance-oriented phrasings in addition to creation-oriented ones. Users who say "tidy up my skill", "review the SKILL.md", "audit this skill", or "clean up the skill directory" will now have the skill loaded and its guidance applied.

Security Impact (required)

New permissions/capabilities? No
Secrets/tokens handling changed? No
New/changed network calls? No
Command/tool execution surface changed? No
Data access scope changed? No

Repro + Verification

Environment

OS: Linux (container, x64)
Runtime/container: OpenClaw gateway (node v22)
Model/provider: anthropic/claude-sonnet-4-6
Integration/channel: N/A (skill metadata only)
Relevant config: N/A

Steps

Before:

Open a session with OpenClaw
Say: "tidy up my pdf-editor skill"
Model proceeds without loading skill-creator skill

After:

Open a session with OpenClaw
Say: "tidy up my pdf-editor skill"
Model matches the expanded description and loads skill-creator skill before responding

Expected

"tidy up a skill", "audit the skill", "review the skill", "clean up the skill", "improve this skill" all trigger the skill-creator skill

Actual (before fix)

Only explicit creation phrases like "create a skill" or "design a skill" triggered it

Evidence

Trace/log snippets

Before (description that was too narrow):

description: Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.

After (description with explicit trigger phrases):

description: Create, edit, improve, or audit AgentSkills. Use when creating a new skill from scratch or when asked to improve, review, audit, tidy up, or clean up an existing skill or SKILL.md file. Also use when editing or restructuring a skill directory (moving files to references/ or scripts/, removing stale content, validating against the AgentSkills spec). Triggers on phrases like "create a skill", "author a skill", "tidy up a skill", "improve this skill", "review the skill", "clean up the skill", "audit the skill".

pnpm build ✅ · pnpm check ✅ (pre-existing tlon/TS errors unrelated to this change)

Human Verification (required)

Verified the diff is exactly one line changed in one file (skills/skill-creator/SKILL.md frontmatter)
Verified the new description includes all phrases mentioned in issue skill-creator: description too narrow — fails to trigger on edit/refactor tasks #36039: "tidy up a skill", "review the skill", "audit the skill", "clean up the skill", "improve this SKILL.md", "edit an existing skill"
Verified pnpm build completes successfully with no new errors
Verified pnpm check output matches pre-existing failures only (tlon module resolution + two ui TS errors present on main before this branch)
What I did not verify: live session trigger testing against a running OpenClaw instance (description matching is purely text-based — the expanded text mechanically subsumes the original)

Compatibility / Migration

Backward compatible? Yes — existing triggers still work; new triggers added only
Config/env changes? No
Migration needed? No

Failure Recovery (if this breaks)

Revert: git revert HEAD on main (single-commit change, clean revert)
Files to restore: skills/skill-creator/SKILL.md — restore the description: line to the original value
Known bad symptoms: None expected; worst case is the description is slightly broader than necessary, which has no harmful effects

Risks and Mitigations

Risk: Overly broad description triggers the skill for unrelated skill-adjacent queries (e.g. "list my skills")
- Mitigation: The trigger phrases are specific action verbs (tidy up, improve, audit, review, clean up) that only apply when a user intends to work on a skill. "List my skills" does not match any of them.

🤖 Generated with Claude Code · This PR was authored with AI assistance (Claude / OpenClaw agent session). The description change was authored, verified, and reviewed by the agent; build and check were run locally before filing.

…iew triggers Closes openclaw#36039 The previous description was too narrow ('Create or update AgentSkills...'), causing the skill to be skipped when users phrase requests as 'tidy up a skill', 'review the skill', 'audit the skill', 'clean up the skill', or 'improve this SKILL.md'. Expand the description to list these trigger phrases explicitly.

greptile-apps · 2026-03-07T20:11:16Z

Greptile Summary

This PR expands the description frontmatter in skills/skill-creator/SKILL.md to cover maintenance-oriented trigger phrases (edit, audit, review, tidy up, clean up) in addition to the original creation-oriented ones. The change is minimal, backward-compatible, and correctly follows the skill's own documented guidance that the description should include all "when to use" context and explicit trigger phrases.

Only one line changed in one file — no logic, scripts, or body content touched.
The new description (~82 words) stays within the skill's own ~100-word metadata guideline (line 117 of the same file).
All six phrases cited in the linked issue are covered: "tidy up a skill", "review the skill", "audit the skill", "clean up the skill", "improve this SKILL.md", and "edit an existing skill".
The inclusion of the explicit Triggers on phrases like … clause aligns with the skill-creator's own frontmatter guidance (lines 324–327), which recommends listing concrete trigger phrases.

Confidence Score: 5/5

Safe to merge — purely additive documentation change with no code, logic, or configuration impact.
The change is a single-line frontmatter edit that only broadens the skill's trigger vocabulary. It is backward-compatible, follows the skill's own design guidelines, and carries no risk of breaking existing behaviour.
No files require special attention.

_{Last reviewed commit: 13997c1}

haynzz · 2026-03-07T20:25:12Z

CI Failure Note

Both failing jobs ( and ) are pre-existing failures on itself and are not caused by this PR (which only touches and ).

Evidence: The most recent CI run on (run #22806401873) fails on the same two jobs:

— falls back to a full scan (shallow clone lacks the base commit) and updates line numbers in the baseline → exits with code 3
— TS2741 error ( property missing) from a recently-merged commit

This PR is unblocked pending those upstream fixes landing on .

haynzz · 2026-03-07T20:25:29Z

CI Failure Note

Both failing jobs (secrets and check) are pre-existing failures on main itself — not caused by this PR, which only touches skills/skill-creator/SKILL.md and .secrets.baseline.

Evidence: The most recent CI run on main (run #22806401873) fails on the same two jobs:

secrets — detect-secrets falls back to a full scan (shallow clone lacks the base commit) and updates line numbers in the baseline → exits with code 3
check — src/cron/isolated-agent/delivery-dispatch.double-announce.test.ts:80 TS2741 error (mode property missing) from a recently-merged commit

This PR is unblocked pending those upstream fixes landing on main.

altaywtf · 2026-03-08T09:13:19Z

Merged via squash.

Prepared head SHA: 13997c1
Merge commit: aedf3ee

Thanks @haynzz.

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

@altaywtf

…iew triggers (openclaw#39158) Merged via squash. Prepared head SHA: 13997c1 Co-authored-by: haynzz <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf

openclaw-barnacle bot added the size: XS label Mar 7, 2026

openclaw-barnacle bot added size: L and removed size: XS labels Mar 7, 2026

haynzz force-pushed the fix/skill-creator-description branch from f7475e4 to 13997c1 Compare March 7, 2026 21:21

openclaw-barnacle bot added size: XS and removed size: L labels Mar 7, 2026

altaywtf merged commit aedf3ee into openclaw:main Mar 8, 2026
43 of 45 checks passed

This was referenced Mar 8, 2026

fix(skill-creator): expand description to cover edit/refactor tasks #36304

Closed

fix(skill-creator): broaden description to trigger on edit/refactor tasks (#36039) #36319

Closed

github-actions bot mentioned this pull request Mar 8, 2026

📡 Upstream Digest — 2026-03-08 10:17 UTC curtismercier/openclaw-mods#209

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(skills): expand skill-creator description to cover edit/audit/review triggers#39158

fix(skills): expand skill-creator description to cover edit/audit/review triggers#39158
altaywtf merged 1 commit intoopenclaw:mainfrom
Nimbl-SaaS:fix/skill-creator-description

haynzz commented Mar 7, 2026

Uh oh!

greptile-apps bot commented Mar 7, 2026

Uh oh!

haynzz commented Mar 7, 2026

Uh oh!

haynzz commented Mar 7, 2026

Uh oh!

Uh oh!

altaywtf commented Mar 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

haynzz commented Mar 7, 2026

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual (before fix)

Evidence

Human Verification (required)

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Uh oh!

greptile-apps bot commented Mar 7, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

haynzz commented Mar 7, 2026

CI Failure Note

Uh oh!

haynzz commented Mar 7, 2026

CI Failure Note

Uh oh!

Uh oh!

altaywtf commented Mar 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants