architektapx/auth-guard/doctor

Claude Code and agent skill by architektapx/auth-guard

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Verify the auth-guard credential protection is fully working. Checks betterleaks and jq availability, hook presence and registration, sandbox and deny-rule settings, which custom-checks file resolves and whether it is valid, and runs redaction and custom-check self-tests with synthetic inputs. Use for "auth-guard doctor", "check auth-guard", "is my secret protection working", "are my custom checks valid".

Install

npx vskill@latest install architektapx/auth-guard/doctor

SKILL.md

auth-guard doctor

Run exactly one command:

${CLAUDE_PLUGIN_ROOT}/bin/auth-guard-doctor.sh

It prints one PASS/WARN/FAIL line per check and exits 0 (healthy, warnings allowed) or 1 (at least one FAIL).

Alongside the dependency, hook and settings checks it covers the user's custom checks: which custom-checks file resolves (or that none does), whether a lower-precedence candidate file is being shadowed by it, whether that file is valid (the guard itself validates and doctor relays its reason), and an end-to-end self-test that runs a synthetic custom-checks file through the guard.

Report to the user:

  • One-line verdict first: healthy, degraded (warnings), or broken (failures).
  • Then only the WARN and FAIL lines with a short plain-language explanation of the consequence of each, and the concrete fix:
  • betterleaks missing: brew install betterleaks (macOS), or see the Dependencies section of the plugin README for Linux - until then tool output is NOT scanned.
  • hook registration dormant: the plugin is installed but not enabled, or hooks were removed from settings; suggest /plugin to enable, or a session restart if it was just enabled.
  • sandbox / deny-rule warnings: suggest running /auth-guard:global-settings.
  • custom-checks file invalid: the whole file is ignored and the guard asks on everything the built-in denies do not already block; report the check name or key the FAIL line names and point the user at docs/custom-checks.md for that field's rules.
  • custom-checks shadowing: only the winning path is read; the others are dead until removed or merged into it.
  • Do not list every PASS line; summarize passes in one sentence.

Never work around a FAIL by editing files yourself; this skill only diagnoses.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/architektapx/auth-guard/doctor/badge)](https://verified-skill.com/skills/architektapx/auth-guard/doctor)

Something wrong with this skill or its page? Tell us.