# Gaf Manifest Doc Consistency Auditor

> Audits consistency between GAF task artifacts and implementation, including task-definition.md, manifest.yaml, automation-doc.md, and task code. Use before coding, during review, and before validation to catch cross-file drift.

- Skill: `miguel-carrera/gaf-manifest-doc-consistency-auditor` (Agent Skill)
- Install (CLI): `npx skillmds@latest add miguel-carrera/gaf-manifest-doc-consistency-auditor`
- Raw SKILL.md: https://api.skillmd.com/api/skills/miguel-carrera/gaf-manifest-doc-consistency-auditor/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: miguel-carrera (https://skillmd.com/u/miguel-carrera)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/miguel-carrera/gaf-manifest-doc-consistency-auditor

---


# GAF Manifest and Doc Consistency Auditor

## Goal
Detect mismatches across task metadata and dependency declarations before they cause runtime failures.

## Use This Skill When
- User asks for a consistency check.
- Preparing to implement or review a task.
- Preparing for Phase 3 validation.

## Files To Compare
- `gaf-automations-library/Test/{task_id}/task-definition.md`
- `gaf-automations-library/Test/{task_id}/manifest.yaml`
- `gaf-automations-library/Test/{task_id}/automation-doc.md`
- `gaf-automations-library/Test/{task_id}/*` runtime files
- `gaf-core-knowledge/task-registry/task-catalog.yaml` (for registry alignment)

## Checks
- `task_id`, name, owner, status, version aligned across files.
- `task-catalog.yaml` matches task artifacts; catalog changes must use `register-task.js` / `deploy-task.js` (read-only comparison here — do not edit catalog by hand).
- `storage_dependencies` in manifest match referenced data sources in code/docs and declared entities in `data-dictionary.yaml`.
- `shared_dependencies` folders exist under `gaf-automations-library/shared` and use **published** paths (`*_v<n.n.n>`) for deploy-ready tasks (not dev-only folder names).
- failure protocol fields align with operational docs and task definition.
- no undeclared entities or storage paths used in implementation.

## Reporting
Return findings as:
- `Critical`: must fix before validation/deploy.
- `Warning`: should fix for maintainability/governance.
- `Info`: optional cleanup.

For each finding include:
- affected file(s)
- mismatch detail
- exact fix recommendation

## Guardrails
- Do not silently rewrite user intent.
- Prefer explicit recommendations over automatic large rewrites.