doc-drift-audit
Skill0 starsChanged 2 months agoMITView on GitHub
Audit whether the product documentation still matches the actual code, finding undocumented endpoints, untested requirements, and direct contradictions between docs and implementation. Use monthly, before a release, or when the user asks "are my docs still accurate", "check for drift", "do the docs match the code", or "audit my documentation".
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- add-featureSkill
- bootstrap-docsSkill
- implement-featureSkill
- release-prepSkill
- reverse-engineer-docsSkill
- stack-auditSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Read AGENTS.md and the full procedure in prompts/05-doc-drift-audit.md.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Documentation Drift Audit
Find every place where the documentation and the code disagree.
Documentation drift is silent: nothing fails, no test goes red — the docs just quietly stop being true, and then an AI session builds against a lie.
Before you start
Read AGENTS.md and the full procedure in prompts/05-doc-drift-audit.md.
Posture
Be adversarial. Assume drift exists — your job is to find it, not to reassure. A clean report that missed something is worse than useless.
Do not fix anything. Report only. The user decides what to act on, and in which direction.
Six audits
| # | Audit | Looking for |
| 1 | Code exists, docs do not | Routes with no SCR-, endpoints with no API-, tables with no ENT-, dependencies with no TECH-, env vars with no CFG- |
| 2 | Docs exist, code does not | Requirements marked Shipped with no implementation, TC- entries with no test |
| 3 | Both exist but disagree | Response shapes, validation rules, field types, status codes, permissions, design tokens, installed versions vs the TECH- register |
| 4 | Traceability integrity | Run all 18 reverse-trace checks from docs/09-traceability-matrix.md §3 |
| 5 | Document hygiene | Unresolved [NEEDS CONFIRMATION] markers, missing change log rows, stale last_updated, terminology inconsistent with the glossary |
| 6 | Safety posture | Branch protection, CI gates, preview/production isolation, secret scanning, backup restore drill, EOL technologies |
Category 3 is the most dangerous — everything looks fine. For each, quote both sides: what the document says, what the code does, with file paths.
Output
- Executive summary — three sentences. Are the docs trustworthy right now, yes or no, and what is the single most important thing to fix?
- Findings register — ID, finding, audit, severity, doc says, code does, evidence, recommended action
- Drift score — counts by category
- Remediation plan — for each finding, state whether the fix is update the document (code is right), change the code (document is right), or a decision the user must make
- Process recommendation — if the same category of drift keeps recurring, the process is at fault, not the people. Say what should change.
Severity: Critical (security exposure, undocumented personal data) · High (will cause someone to build the wrong thing) · Medium (wastes time) · Low (cosmetic).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

