Agent skill
doc-health-audit
Quellcode ansehen: nickcrew/claude-cortex
Installation
npx skills add nickcrew/claude-cortex --skill doc-health-audit51
Installationen
EU-hosted inference API
Power your AI agent skills with open-source models.
Drop-in OpenAI-compatible API. No data leaves Europe.
MiniMax
MiniMax M3
$0.50 / $1.90
per M tokens
Z.ai
GLM 5.3 Flash
$0.20 / $0.60
per M tokens
MoonshotAI
Kimi K3
$4.00 / $18.00
per M tokens
DeepSeek
DeepSeek V4.1 Flash
$0.40 / $1.42
per M tokens
Documentation Health Audit
Orchestrate a comprehensive documentation assessment across five dimensions.
Each phase builds on the previous — structural issues are found before accuracy
is checked, accuracy before completeness, and so on.
This skill does not duplicate any phase logic. It defines the run order, phase
gating, and combined summary. Each phase delegates to its dedicated skill.
When to Use
- Pre-release audit — full pass before shipping
- Periodic health check — monthly or quarterly
- After a major refactor that touched many docs
- When documentation quality is unknown or untrusted
- First-time assessment of a doc set
For single-dimension checks, use the individual skills directly.
Pipeline
Phase 1: Structure → doc-maintenance
Phase 2: Accuracy → doc-claim-validator
Phase 3: Completeness → doc-completeness-audit
Phase 4: Quality → doc-quality-review
Phase 5: Architecture → doc-architecture-reviewWhy This Order
Each phase depends on the prior phase being reasonably healthy:
| Phase | Depends on | Reason |
|---|---|---|
| Structure | — | Foundation. Broken links and orphans must be visible before anything else |
| Accuracy | Structure | Can't verify claims in docs you can't find or that have broken references |
| Completeness | Accuracy | No point inventorying gaps if existing docs contain false claims |
| Quality | Completeness | Review prose quality only on docs you intend to keep |
| Architecture | Completeness | Need to know what's missing before evaluating whether the structure can hold it |
Execution
Step 1: Determine Scope
Before running phases, establish what to audit:
- Full audit — all documentation (
docs/,site/,README.md) - Scoped audit — specific directory or doc set (e.g., "just the reference section")
- Changed-only audit — docs modified since a git ref (
git diff --name-only main...HEAD -- '*.md')
Pass the scope to each phase so they operate on the same set.
Step 2: Run Phases with Gating
Run each phase sequentially. After each phase, evaluate the gate condition.
Phase 1: Structure (doc-maintenance)
Run: Invoke doc-maintenance on the scoped doc set.
Gate condition: Check the findings.
- If >10 broken internal links → WARN the user: "Structural issues found. Fix these
first for accurate results in later phases, or continue anyway?" - If critical structural issues (missing index pages, completely broken navigation) →
RECOMMEND STOP. Later phases will produce unreliable results. - Otherwise → CONTINUE
Phase 2: Accuracy (doc-claim-validator)
Run: Invoke doc-claim-validator on the scoped doc set.
Gate condition: Check the findings.
- If P0 findings (user-facing docs that would break if followed) → WARN the user.
These should be fixed before assessing completeness, but the audit can continue. - Otherwise → CONTINUE
Phase 3: Completeness (doc-completeness-audit)
Run: Invoke doc-completeness-audit on the scoped doc set.
Gate condition: No gate — always continue to quality. Completeness gaps don't
affect quality assessment of existing docs.
Phase 4: Quality (doc-quality-review)
Run: Invoke doc-quality-review on the scoped doc set (excluding any docs
flagged for removal in earlier phases).
Gate condition: No gate — always continue to architecture.
Phase 5: Architecture (doc-architecture-review)
Run: Invoke doc-architecture-review on the scoped doc set. Pass the
completeness gap list from Phase 3 so the architecture review can assess whether
the current structure can accommodate the missing docs.
Gate condition: Final phase — no gate.
Step 3: Produce Combined Summary
After all phases complete (or after an early stop), merge results into a single
dashboard.
Combined Summary Format
# Documentation Health Audit
**Audit date:** YYYY-MM-DD
**Scope:** [what was audited]
**Phases completed:** N / 5
**Early stop:** [yes/no — if yes, which phase and why]
---
## Dashboard
| Phase | Skill | Status | Key Metric | Top Finding |
|-------|-------|--------|------------|-------------|
| 1. Structure | doc-maintenance | PASS/WARN/FAIL | N broken links, N orphans | [one line] |
| 2. Accuracy | doc-claim-validator | PASS/WARN/FAIL | N P0, N P1 claims failed | [one line] |
| 3. Completeness | doc-completeness-audit | PASS/WARN/FAIL | N% coverage, N P0 gaps | [one line] |
| 4. Quality | doc-quality-review | PASS/WARN/FAIL | Grade: [A-F], avg score N/25 | [one line] |
| 5. Architecture | doc-architecture-review | PASS/WARN/FAIL | Grade: [A-F], N/35 score | [one line] |
**Overall health:** [Healthy / Needs Attention / Critical]
---
## Priority Actions
Top 5 actions across all phases, ordered by impact:
1. [Phase N] [specific action] — [why it matters]
2. [Phase N] [specific action] — [why it matters]
3. [Phase N] [specific action] — [why it matters]
4. [Phase N] [specific action] — [why it matters]
5. [Phase N] [specific action] — [why it matters]
---
## Phase Reports
[Include or link to each phase's full report below]
### Phase 1: Structure
[doc-maintenance report]
### Phase 2: Accuracy
[doc-claim-validator report]
### Phase 3: Completeness
[doc-completeness-audit report]
### Phase 4: Quality
[doc-quality-review report]
### Phase 5: Architecture
[doc-architecture-review report]Partial Runs
Not every audit needs all five phases. Common partial runs:
| Scenario | Phases to run |
|---|---|
| "Are the docs accurate?" | 1 + 2 only |
| "What's missing?" | 1 + 3 only |
| "Is the writing good?" | 4 only |
| "Can users find things?" | 1 + 5 only |
| "Quick pre-release check" | 1 + 2 + 3 (skip quality and architecture) |
| "Full audit" | All 5 |
When running a subset, skip gates for omitted phases.
Anti-Patterns
- Do not run all five phases if you only need one — use the individual skill directly
- Do not skip Phase 1 (structure) — it's the foundation for everything else
- Do not ignore gate warnings — they exist because later phases produce unreliable results on broken foundations
- Do not combine this with doc remediation in the same session — audit first, fix separately
- Do not run this on archived docs (
docs/archive/) — they are historical
Verwandte Skills
Mehr aus dieser Quelle: nickcrew/claude-cortex
react-performance-optimization
nickcrew/claude-cortex
1.9K installs
owasp-top-10
nickcrew/claude-cortex
760 installs
ui-design-aesthetics
nickcrew/claude-cortex
362 installs
threat-modeling-techniques
nickcrew/claude-cortex
357 installs
token-efficiency
nickcrew/claude-cortex
245 installs
repo-cleanup
nickcrew/claude-cortex
183 installs
So verwenden Sie diesen Skill
- 1
Install doc-health-audit by running
npx skills add nickcrew/claude-cortex --skill doc-health-auditin your project directory. Führen Sie den obigen Installationsbefehl in Ihrem Projektverzeichnis aus. Die Skill-Datei wird von GitHub heruntergeladen und in Ihrem Projekt platziert. - 2
Keine Konfiguration erforderlich. Ihr KI-Agent (Claude Code, Cursor, Windsurf usw.) erkennt installierte Skills automatisch und nutzt sie als Kontext bei der Code-Generierung.
- 3
Der Skill verbessert das Verständnis Ihres Agenten für doc-health-audit, und hilft ihm, etablierte Muster zu befolgen, häufige Fehler zu vermeiden und produktionsreifen Code zu erzeugen.
Was Sie erhalten
Skills sind Klartext-Anweisungsdateien — kein ausführbarer Code. Sie kodieren Expertenwissen über Frameworks, Sprachen oder Tools, das Ihr KI-Agent liest, um seine Ausgabe zu verbessern. Das bedeutet null Laufzeit-Overhead, keine Abhängigkeitskonflikte und volle Transparenz: Sie können jede Anweisung vor der Installation lesen und prüfen.
Kompatibilität
Dieser Skill funktioniert mit jedem KI-Coding-Agenten, der das skills.sh-Format unterstützt, einschließlich Claude Code (Anthropic), Cursor, Windsurf, Cline, Aider und anderen Tools, die projektbezogene Kontextdateien lesen. Skills sind auf Transportebene framework-agnostisch — der Inhalt bestimmt, für welche Sprache oder welches Framework er gilt.
Data sourced from the skills.sh registry and GitHub. Install counts and security audits are updated regularly.