Skill Review.

You are an expert reviewer of Agent Skills (SKILL.md files). Perform a structured quality review of the SKILL.md below.

SKILL_MD_CONTENT:
{{SKILL_MD_CONTENT}}

Review the SKILL.md against each of the following criteria. For each criterion, give a rating (Pass / Needs work / Fail) and a one-sentence justification. Where the rating is not Pass, provide a specific, actionable fix.

**1. Frontmatter completeness**
- Does it have name (kebab-case), description, version, and metadata?
- Is name unique and descriptive?

**2. Description quality (most important)**
- Is the description 2–5 sentences?
- Does it state what the skill produces specifically (not just a category)?
- Does it include explicit trigger language ("Use when asked to...")?
- Does it include explicit exclusion language ("Do NOT use for...")?
- Would a QA engineer be confident the agent will activate this at the right time?

**3. Trigger scope**
- Is the description specific enough to avoid false positives?
- Is it broad enough to activate for paraphrased requests, not just exact phrases?

**4. Instruction body completeness**
- Are there clearly labelled sections for: When to use, Inputs, Instructions, Output format?
- Are instructions numbered and sequential?
- Is each instruction atomic (one action per step)?
- Does the output format section describe the exact file name(s), structure, and any follow-up summary?

**5. Inputs section**
- Are all required inputs listed with descriptions?
- Are optional inputs distinguished from required ones?

**6. Anti-patterns and safety**
- Does it have an Anti-patterns section with at least 2 specific items?
- Does it have a Safety section that prohibits credentials/secrets in generated output?
- Does it state that generated output is a draft requiring review?

**7. Maintainability**
- Is inline content (coding conventions, style guides, large reference material) a candidate for extraction to references/?
- Is the version field present and does the metadata include a last-reviewed date?

**8. Overall assessment**
After the per-criterion review, provide:
- An overall quality rating: Production-ready / Needs minor revision / Needs major revision
- The top 3 most impactful changes to make
- One optional enhancement that would make this skill significantly more useful

You are an expert in Agent Skill activation and SKILL.md description authoring. Diagnose and fix the description field below.

CURRENT_DESCRIPTION:
{{CURRENT_DESCRIPTION}}

ACTIVATION_PROBLEM: {{ACTIVATION_PROBLEM}}
(e.g. "false positives — agent activates for unit test requests", "false negatives — agent misses requests to generate Playwright tests", "both")

SKILL_ACTUAL_PURPOSE:
{{SKILL_ACTUAL_PURPOSE}}

GOOD_ACTIVATION_EXAMPLES:
{{GOOD_ACTIVATION_EXAMPLES}}
(3 example user requests that SHOULD activate this skill)

BAD_ACTIVATION_EXAMPLES:
{{BAD_ACTIVATION_EXAMPLES}}
(3 example user requests that should NOT activate this skill)

Provide:

**1. Diagnosis**
What specifically in the current description causes the stated problem? Be precise — quote the problematic phrase(s).

**2. Rewritten description**
Apply these rules:
- Sentence 1: What this skill produces (specific output, not category)
- Sentence 2: "Use when [specific trigger phrases from GOOD_ACTIVATION_EXAMPLES]"
- Sentence 3: "Do NOT use for [explicit exclusions covering BAD_ACTIVATION_EXAMPLES]"
- Total: 2–5 sentences

**3. Validation**
For each of the 6 example requests (3 good, 3 bad), state whether the new description would now correctly activate or exclude the skill, and why.

**4. Edge cases**
Identify 2 edge-case requests not in the examples above where the correct activation behaviour is ambiguous. For each, recommend whether to activate or exclude, and suggest how to adjust the description to handle it.

You are an expert in Agent Skill design. Audit the trigger scope of the SKILL.md(s) below to identify whether they are too broad, too narrow, or correctly calibrated.

SKILLS_TO_AUDIT:
{{SKILLS_TO_AUDIT}}
(Paste one or more SKILL.md description fields and "When to use" sections, labelled by skill name)

ADJACENT_SKILLS:
{{ADJACENT_SKILLS}}
(Paste the description fields of any skills that might overlap with the ones being audited — or write "none")

INTENDED_COVERAGE:
{{INTENDED_COVERAGE}}
(Describe the full range of user requests this set of skills should collectively handle)

Provide:

**1. Per-skill scope rating**
For each skill: Correctly calibrated / Too broad / Too narrow / Conflicts with adjacent skill
With a one-sentence justification and a specific fix.

**2. Overlap analysis**
Are any two skills likely to both activate for the same user request? List the overlapping scenarios and recommend which skill should "win" (and how to adjust the losing skill's description to yield correctly).

**3. Gap analysis**
Are there user requests within INTENDED_COVERAGE that no skill in the set would activate for? List the gaps and recommend either: adjusting an existing skill's scope to cover the gap, or creating a new skill.

**4. Recommended description adjustments**
For each skill that needs adjustment, provide the specific rewritten description (2–5 sentences, "Use when" + "Do NOT use for" structure).

You are an expert in Agent Skill design and refactoring. Analyse the SKILL.md below, identify natural split points, and produce focused sub-skills.

MONOLITHIC_SKILL_MD:
{{MONOLITHIC_SKILL_MD}}

SPLIT_REASON: {{SPLIT_REASON}}
(e.g. "The skill handles both test generation and test review — they have different inputs, outputs, and trigger phrases")

TARGET_AGENT: {{TARGET_AGENT}}

Provide:

**1. Split analysis**
How many sub-skills should this be split into? List each proposed sub-skill with:
- Its name (kebab-case)
- The subset of the original instructions it handles
- The trigger phrases it should own
- The workflows it explicitly excludes from the original skill

**2. Sub-skill SKILL.md files**
For each sub-skill, output a complete SKILL.md in a fenced code block:
- Frontmatter with name, description (2–5 sentences, trigger + exclusion language), version, metadata
- Full instruction body derived from the relevant subset of the original skill

**3. Overlap check**
After generating the sub-skills, verify: are there any user requests that would activate two sub-skills simultaneously? If yes, adjust the descriptions to resolve the overlap.

**4. Migration note**
What should be done with the original monolithic skill? (Delete it, archive it, or keep it as an alias that routes to the sub-skills?)