Audit a project verification Skill against source and live user paths, correcting proven drift in its own files.
Pro scans all 2 files and shows the line behind each finding
Scanned 10/1/2026
npx -y skills add williamwue/oh-my-stack --skill maintain-verification-skill --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Maintain Verification Skill?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/williamwue-maintain-verification-skill-d3b8ffbc)More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.
---
name: maintain-verification-skill
description: Audit a project verification Skill against source and live user paths, correcting proven drift in its own files.
---
# Maintain verification Skill
Locate the project verification Skill with launch/drive instructions and a
feature map. If several match, resolve the target from the request or ask.
If none exists, use [create-verification-skill](../create-verification-skill/SKILL.md).
Only edit that Skill, its feature map, and its owned helpers. Product defects
are reported separately; documentation must not hide them.
Check the map index against feature files for missing, duplicate, or dead
entries. For each feature, assign a bounded read-only source pass that returns
the entry points, likely drift with citations, and one live verification recipe.
Parallelize within runtime capacity; sequential root passes are an explicit
fallback. Reconcile every feature and inspect recent source changes for missing
user-facing paths.
The coordinator then runs one live pass covering every feature. Follow the
verification Skill's launch model: one checked shared instance or fresh
isolated sessions. Doctor before first drive and after a surprising failure;
reset a wedged state rather than repeatedly trusting a healthy process check.
For each drive, preserve evidence before cleanup. Clean residual state after
failed attempts and tear down the instance after the final pass. Record an
unreachable feature only with the attempted path and concrete prerequisite.
Classify findings: doc drift, harness gap, or product gap. Fix only the first
two under the owned Skill directory and re-run any changed harness live.
Re-read the final diff and evidence. Outcome is `clean` when all mapped features
have source and live coverage with no correction; `changed` when corrections
are proven; `blocked` when coverage or safe correction cannot finish. A source
scan alone cannot yield `clean`. Return per-feature coverage and remaining
prerequisites. If the user explicitly requested a pull request, create at most
one from the verified changes; otherwise leave the scoped changes local.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!