SkillVaultskills Browse all 1,000+ skills

Maintenance · Version 1.3.0 · Reviewed 2026-08-02

Documentation Decay Auditor

Reduce change risk for staleness detection and impact ranking with evidence, explicit trade-offs, and a verification plan.

4 method steps 5 documented failure modes 5 diagnostic checks 7 quality gates

Finds documentation that has silently diverged from the system and prioritizes what to fix or delete.

₹299 one-time

Get this skill archive

Install in your AI coding tool

SkillVault packages this skill in the open Agent Skills format for five leading coding tools, with a raw SKILL.md fallback for every other compatible IDE or agent.

See the complete graphical installation and usage guide

What this skill helps you do

  • Staleness detection
  • Impact ranking
  • Deletion decisions

How Documentation Decay Auditor works

You provide

The reader task, current pages, source evidence, navigation, and support or search signals

It inspects

Correctness, findability, and task completion for staleness detection

It decides

A impact ranking change with ownership and a refresh trigger

You verify

A clean reader completes the task and the documentation build validates links, metadata, and examples

What it checks first

Documentation Decay Auditor finds documentation that has silently diverged from the system and prioritizes what to fix or delete. Use it when the work involves Staleness detection, Impact ranking, Deletion decisions.

  1. The user task the document must enable and the observable completion condition, rather than the page type alone.
  2. Technical claims traced to current code, configuration, interfaces, or an accountable decision owner.
  3. Navigation and search language aligned with how readers describe the problem instead of the owning team structure.
  4. Executable examples tested in a clean environment with the same versions and prerequisites the page declares.
  5. Ownership, last-reviewed evidence, and a refresh trigger tied to the source behavior most likely to change.

Failure modes it recognizes

  • A correct page remains undiscoverable because its title and navigation use internal vocabulary readers never search.
  • Setup instructions omit an implicit credential, platform, or working-directory assumption and fail on the first command.
  • Documentation describes the intended architecture while production code follows a later undocumented path.
  • Generated navigation excludes a valid page or keeps a deleted page reachable through stale links.
  • A support answer is copied into documentation without verifying that the workaround remains safe and current.

Answers it will reject

  • Measuring documentation quality by page count, word count, or formatting compliance instead of task success.
  • Mirroring the repository tree as navigation when readers approach the system by goals and failures.
  • Publishing code samples that were syntax-checked but never executed against the stated prerequisites.
  • Adding an FAQ entry for every support question instead of repairing the earliest missing concept or task step.

Decision rules it applies

  • Fix incorrect and task-blocking content before improving completeness, style, or visual polish.
  • Prefer one maintained source for a fact and link to it rather than duplicating volatile instructions across pages.
  • Create a new page only when it owns a distinct reader task or decision and has a durable maintenance owner.
  • Retire content when no current reader journey depends on it or when a maintained source supersedes it completely.

Evidence it asks for

  • Run links, metadata, navigation generation, and representative code examples in the documentation build.
  • Trace support and search queries to the page and section where the reader task currently fails.
  • Test the highest-value journeys from a clean environment using only the published instructions.
  • Diff documented commands, interfaces, and configuration against their defining source at release time.

The method inside

  1. Map the artifact, actors, boundaries, and invariants relevant to staleness detection.
  2. Trace concrete failure or abuse paths for impact ranking; do not report checklist items without a mechanism.
  3. Prioritize deletion decisions findings by impact, likelihood, confidence, and cost of correction.
  4. Recommend the smallest defensible change, then define how an independent reviewer can verify it.

Deliverables

  • Staleness detection assessment
  • Impact ranking decision and action plan
  • Deletion decisions verification checklist

Evidence requirements

  • Current and target versions
  • Dependency graph and changelogs
  • Tests, compatibility constraints, and rollout environment

Quality gates

  • Every material claim traces to supplied evidence or is labeled as a hypothesis.
  • The response follows the declared deliverable contract.
  • No execution, access, measurement, or verification is invented.
  • Secrets and personal data are redacted rather than repeated.
  • The user receives a concrete independent verification step.
  • The relevant failure modes in this domain were considered rather than only the reported symptom.
  • No listed anti-pattern was recommended as a solution.

Example task

Input

Our internal docs are large and partially wrong, and people have stopped trusting any of it.

Expected output

Partial correctness is worse than absence because readers cannot tell which parts to trust. Rank by traffic and consequence, verify the top pages against the system, and delete rather than preserve anything unverifiable, since deletion restores trust faster than rewriting...

Boundaries and compatibility

Ideal for

  • Staleness detection: produce a decision or artifact grounded in supplied evidence.
  • Impact ranking: produce a decision or artifact grounded in supplied evidence.
  • Deletion decisions: produce a decision or artifact grounded in supplied evidence.

Out of scope

  • Blindly upgrading across multiple major versions
  • Assuming semantic versioning guarantees compatibility

Agent compatibility

  • GitHub Copilot Agent Skills
  • Cursor Agent Skills
  • Claude Code Skills
  • OpenAI Codex Skills
  • JetBrains Junie Skills

Tool policy: Advisory by default. No tools are assumed. If the host provides tools, use read-only evidence gathering unless the user explicitly approves a scoped write or execution action.