Review this vault. Find what is broken or disconnected, report it in plain language, and give the exact call that fixes each one. Writes nothing. Use for "what's wrong here" · "check the vault" · "is this healthy".
Scanned 9/2/2026
Install to Claude Code
npx -y skills add wlsdks/ontology-atlas --skill atlas-review --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Atlas Review?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/wlsdks-atlas-review-ontology-atlas)More formats (shields.io, HTML) on the badges page.
---
name: atlas-review
description: Review this vault. Find what is broken or disconnected, report it in plain language, and give the exact call that fixes each one. Writes nothing. Use for "what's wrong here" · "check the vault" · "is this healthy".
---
# /atlas-review — is this vault sound right now
**This skill writes nothing.** What it produces is one page: what is off, and
what call fixes it. Fixing happens after a person has read it and decided.
1. **Two calls**
- `validate_vault({})` — frontmatter and relation references
- `query_ontology({ operation: 'health' })` — the graph integrity checks
Do not `grep` or `sed` the frontmatter directly. You get the same answer more
slowly, without relation resolution or schema validation.
2. **Do not report what passed**
Listing healthy checks spends the reader's attention and buries the one thing
they need to act on. Report **only what needs a hand**. If everything is sound,
end in one line — "nothing to fix · N nodes · M relations".
3. **Three things per item**
| | |
|---|---|
| What happened | One plain sentence. Do not paste the check's identifier |
| Why it matters | What goes missing or renders wrong if it stays |
| The fix | One line the reader can copy and run |
The two you will see most:
- **Unreachable nodes** (`components`) — a node the project root cannot reach
belongs to no project, so it never appears on the map. It exists and is invisible.
→ `add_relation('<project>', 'domains/<new domain>', 'domains')`
- **One-sided relations** (`relation_recommendations`) — a capability declares
`domain: X` but `X` does not claim it back.
→ `add_relation('domains/X', 'capabilities/<capability>', 'capabilities')`
4. **If project meaning is yellow — stop here**
Never finalize `meaning_assessment` without human approval. The project's five
competency answers are a claim about what this project *is*, and an agent
cannot settle that on someone's behalf. Show what is missing and stop.
## How this skill fails
- Lists passing checks until the actionable one is buried
- Pastes the check identifier without translating what it means
- States the problem with no fix line — so the reader has to ask again
- Fixes things without asking
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!