---
name: system-diagnostic
description: Audits the robot assistant system for consistency and health. Use this skill whenever the user asks to check system health, audit skills, run a diagnostic, verify the skill index, or find redundant/missing/outdated skills. Also trigger proactively after major changes like adding or removing multiple skills.
---

# System Diagnostic

Perform a structured audit of the robot assistant system and produce a health report.

## When to Run

- User explicitly asks for a diagnostic, audit, or health check
- After adding or removing several skills at once
- When the system feels like it's drifted — skills not triggering right, index feels stale, etc.

---

## Audit Checklist

Work through each check in order. Collect all findings before writing the report — don't stop at the first issue.

### 1. Index ↔ Folder Parity

- List every entry in `skills/_skill-index.md`
- List every folder in `skills/` that contains a `SKILL.md`
- Flag any index entry with no corresponding folder (ghost entry)
- Flag any folder with no index entry (orphaned skill)

### 2. Trigger Collision Check

- Collect all trigger phrases from the Quick Trigger Reference table in the index
- Check for exact matches or near-matches across skills (e.g., two skills that would both fire on "create a task")
- Flag any collisions or ambiguous overlaps

### 3. Redundancy Check

- Read the `description` frontmatter of each `SKILL.md`
- Identify any two skills with substantially overlapping purpose — skills that would handle the same request
- Flag pairs where one skill likely makes the other unnecessary, or where they'd conflict

### 4. AGENTS.md Health

Check that AGENTS.md contains:
- [ ] Identity section (name, role)
- [ ] Preferences
- [ ] Rules
- [ ] Reference to skill index
- [ ] Connected Tools section (even if empty)
- [ ] Workspace structure notes

### 5. Skill File Quality

For each `SKILL.md`, verify:
- [ ] Has `name` frontmatter field
- [ ] Has `description` frontmatter field (this is the primary trigger mechanism — missing or vague descriptions cause skills to not fire)
- [ ] Body has at least minimal instructions (not just a stub)

### 6. Dependencies Check

For any skill that lists dependencies (tools, CLIs, MCP connectors):
- Note whether those dependencies are confirmed available

---

## Report Format

ALWAYS use this structure:

```
# System Health Report
Generated: [date]

## Summary
One-paragraph overview of overall system state — healthy, needs attention, or has critical gaps.

## Passing Checks
- [List what's clean and working]

## Issues Found
### [Issue Title]
**Type:** Ghost entry / Orphan / Collision / Redundancy / Missing field / etc.
**Details:** What exactly is wrong and where
**Suggested fix:** What to do about it

[Repeat for each issue]

## Recommendations
- [Next skill to build, based on gaps or the user's work context]
- [Any maintenance actions worth doing soon]

## Maintenance Checklist
Things to verify whenever the system grows:
- [ ] Every skill folder has a SKILL.md
- [ ] Every skill in the index has a folder
- [ ] No trigger phrase collisions
- [ ] Build log reflects current skill set
- [ ] AGENTS.md Connected Tools section is current
- [ ] Skill descriptions match what the skills actually do
```

---

## Output

Save the report as `system-health-report.md` in the workspace root (overwrite if one already exists). Then summarize findings in the conversation — focus on issues and recommendations, not the full report.