| name | syner-enhance-skills |
| description | Improve an existing skill by auditing and applying fixes. Use when asked to "enhance skill", "improve skill", "fix skill issues", or when code-reviewer detects skill file changes. Delegates to syner-skill-reviewer for audit, then proposes and applies fixes with user confirmation. |
| agent | dev |
| metadata | {"author":"syner","version":"0.2.0"} |
| tools | ["Read","Glob","Grep","Edit","AskUserQuestion","Skill"] |
Skill Enhancer
Audit a skill and apply fixes with user confirmation.
This skill wraps /syner-skill-reviewer (audit-only) and adds the fix layer. Review reports, you decide. Enhance applies.
Process
- Invoke audit: Run
/syner-skill-reviewer on the target skill
- Parse findings: Extract issues by severity (Critical, Warning, Suggestion)
- Propose fixes: For each finding, generate a concrete fix
- Confirm with user: Use
AskUserQuestion to confirm changes
- Apply fixes: Use
Edit to apply approved changes
Input
The skill name or path. Examples:
syner-skill-reviewerskills/syner/SKILL.md.claude/skills/my-skill/SKILL.md
If no skill specified, use AskUserQuestion to ask which skill to enhance.
Step 1: Audit
Invoke the reviewer skill:
Skill(skill="syner-skill-reviewer", args="<skill-name> deep")
Always request "deep" review to get all findings.
Step 2: Parse and Prioritize
From the audit output, extract:
- Critical issues — must fix (identity confusion, injection risk)
- Warning issues — should fix (vague tools, fragile paths)
- Suggestion issues — nice to fix (conventions, frontmatter)
Step 3: Generate Fixes
For each finding, determine the concrete change:
| Finding Type | Fix Approach |
|---|
| First-person voice (A1) | Rewrite to imperative form |
| Identity patterns (A2) | Remove "You are", use directives |
| Heading tone (A3) | Rename to descriptive heading |
| Tool specificity (B1) | Add explicit tool names |
| Path resolution (B2) | Anchor to project root |
| Skill references (B3) | Use /skill-name format |
| Missing input handling (B4) | Add default or AskUserQuestion |
| Output structure (B5) | Describe output qualities, remove rigid placeholder templates |
| Description quality (B6) | Replace trigger lists with value-describing sentences |
| Missing delegation (B7) | Add execution contract for multi-step operations |
Step 4: Confirm Changes
Present all proposed changes to the user:
AskUserQuestion with options:
- "Apply all fixes" (recommended)
- "Review each fix individually"
- "Apply critical only"
- "Cancel"
If user chooses individual review, present each fix one at a time.
Step 5: Apply Fixes
Use Edit to apply each approved change. After all edits:
- Read the modified file to verify changes
- Optionally re-run
/syner-skill-reviewer to confirm issues resolved
Output
## Enhanced: `[skill-name]`
### Applied
- [A1] Rewrote "What I Do" → "Process"
- [B2] Anchored paths to project root
### Skipped (user choice)
- [C2] Frontmatter additions
### Verification
Re-audit shows: [X issues remaining | clean]
Edge Cases
- No findings: Report skill is clean, no changes needed
- User cancels all: Acknowledge and exit without changes
- Partial apply fails: Report which edits failed, leave file in consistent state
Boundaries
Validate against /syner-boundaries:
- Route, Don't Hoard — Delegate audit to
/syner-skill-reviewer
- Suggest, Don't Enforce — Confirm before applying fixes
- Self-Verification — Verify changes were applied
- Observable Work — Report what was changed
