# Knowledge Schema Audit

> Use when non-task knowledge needs a read-only check for schema, frontmatter, localization, links, or consistency.

- **Type:** Skill
- **Install:** `agentstack add skill-choral-io-choral-skills-knowledge-schema-audit`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [choral-io](https://agentstack.voostack.com/s/choral-io)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [choral-io](https://github.com/choral-io)
- **Source:** https://github.com/choral-io/choral-skills/tree/main/skills/knowledge-schema-audit
- **Website:** https://github.com/choral-io/choral-skills

## Install

```sh
agentstack add skill-choral-io-choral-skills-knowledge-schema-audit
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Knowledge Schema Audit

## Runtime Context

Before acting, resolve `` using the runtime bootstrap rules, then read `/.workflow/runtime.md` and `/.workflow/manifest.yml`; do not assume non-default workflow paths or default ids.

Use this skill to inspect non-task knowledge schema quality. This skill is read-only.

## Workflow

1. Read `/.workflow/schemas/common.md`.
2. Read relevant area schemas under `/.workflow/schemas/`.
3. Scan `/**/*.md`.
4. Exclude workflow infrastructure: `/.workflow/**` and `/.feedback/**`.
5. Exclude delivery surfaces: `/tasks/**` and `/planning/KANBAN.md`; task metadata, readiness, dependencies, and Kanban consistency belong to `task-metadata-audit`.
6. Exclude local-only state: `/workspace/*/local/**`.
7. Parse frontmatter, filenames, links, and localized-file suffixes.
8. Compare each document to the relevant area schema.
9. Report findings and dry-run fixes without editing files.

## Checks

- Missing YAML frontmatter.
- Missing or invalid `scope`, `type`, `owners`, or `tags`. Treat `owners: []` as valid syntax and report it as an ownership gap only when the document type, workflow state, or requested check requires an accountable owner.
- Ownership metadata that does not use the canonical `owners` field.
- Localized files missing `lang`, `canonical`, or `translation_of`.
- Localized files that link to other localized files by default.
- Discovery, product, user story, design, test case, metric, experiment, release, or guideline files placed in the wrong area.
- Assets referenced from Markdown but not stored under a typed `/assets///` directory.
- Decision files with missing supersession metadata when a replacement is obvious.
- Proposal files with missing or invalid `proposal_type`, `proposal_status`, `sources`, or target metadata.
- User story files with invalid `type`, `status`, actor, relationship, or acceptance-intent metadata.
- Test case files with invalid `type`, `status`, `priority`, `automation`, or coverage metadata.
- Metric files with invalid `type`, `status`, source, unit, target, or review-cadence metadata.
- Experiment files with invalid `type`, `status`, hypothesis, metric, guardrail, or result metadata.
- Release files with invalid `status`, version, date, validation, rollout, rollback, or follow-up metadata.
- Member files missing `member_id` or `display_name`.
- Member files that still use legacy `groups` frontmatter instead of group documents' `members` lists.
- Group `members` entries that point to missing or ambiguous member profiles.
- Workspace notes that appear to contain project facts that should be promoted.
- Shared workspace files created under deprecated `daily/`, `inbox/`, `scratch/`, or `drafts/` directories.
- Possible secrets, credentials, private customer data, or private personal notes.

## Output

Use the standard audit output shape from `references/report.md`: `Summary`, `Findings`, `Dry-run fixes`, `Requires judgment`, `Safe after approval`, and `Sources`.

## Guardrails

- Do not edit files.
- Do not delete or archive documents.
- Do not rewrite localized content.
- Do not infer product intent, design intent, or architecture decisions.
- Stop and report possible sensitive content.

## References

- For report format and fix rules, read `references/report.md`.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [choral-io](https://github.com/choral-io)
- **Source:** [choral-io/choral-skills](https://github.com/choral-io/choral-skills)
- **License:** MIT
- **Homepage:** https://github.com/choral-io/choral-skills

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-choral-io-choral-skills-knowledge-schema-audit
- Seller: https://agentstack.voostack.com/s/choral-io
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.