HookAudit

Skill mj-deving/pai-skills/skills/Beads/HookAudit

Curated, sanitized export of 21 agent-skill packages for Claude Code and Codex, gated by an automated publication audit (no secrets, no local paths).

Install

npx -y skills add mj-deving/pai-skills --skill HookAudit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Audit and repair Claude-side Beads hook behavior on the local machine. USE WHEN Beads hooks, SessionStart, PreCompact, Stop, PostToolUse, hook doctrine, auto-close risk, or machine-level Beads alignment.

SKILL.md

3.1 KB, as published. Nobody here has run it

Beads Hook Audit

Use this skill for machine-level Claude hook behavior, not repo-local workflow files.

Outcome rule:

using this skill should result in machine-level Beads hook hygiene, not a residual checklist of hook commands for another agent to run later
audit, repair, and verify directly whenever the environment permits it
only leave manual follow-up when blocked by permissions or by an explicit user decision to defer repair

Doctrine source:

${HOME}/.claude/PAI/BEADS_ADOPTION.md
references/hook-policy.md

Scope

Audit and repair:

${AGENT_HOME}/settings.json
hook files referenced there
Beads-related SessionStart, PreCompact, Stop, and PostToolUse behavior

Do not rewrite repo CLAUDE.md / AGENTS.md unless explicitly asked.

Quick start

sed -n '1,260p' ${AGENT_HOME}/settings.json
grep -nE 'SessionStart|PreCompact|Stop|PostToolUse|bd prime|Beads|handoff|MEMORY' ${AGENT_HOME}/settings.json
sed -n '1,240p' ${AGENT_HOME}/hooks/BeadsAutoClose.hook.ts
sed -n '1,260p' ${AGENT_HOME}/hooks/LoadContext.hook.ts

If the machine uses different hook files, inspect those instead.

Ask only if needed

At most:

audit only or audit + repair?
Should local continuity files remain as convenience, or be minimized aggressively?

Workflow

1. Enumerate hook points

Check whether the machine uses:

SessionStart
PreCompact
Stop
PostToolUse

Look for:

bd prime
context-loading hooks
auto-close hooks
handoff / memory injection
wrapup helpers

2. Classify behavior

Use the hook policy reference and label each relevant behavior:

good
acceptable but noisy
risky
forbidden

3. Repair only what violates doctrine

Good defaults:

SessionStart: bd prime
PreCompact: bd prime --stealth
local continuity loading only as convenience

Forbidden:

auto-close on commit
auto-close on session end
local next-step queues injected as authoritative startup context
repo-specific workflow assumptions encoded globally

Prefer weakening risky hooks into reminders instead of deleting useful behavior blindly.

Validation

grep -nE 'SessionStart|PreCompact|Stop|PostToolUse|bd prime|Beads|handoff|MEMORY' ${AGENT_HOME}/settings.json
sed -n '1,240p' ${AGENT_HOME}/hooks/BeadsAutoClose.hook.ts
grep -nE 'Next steps|Handoff|Local handoff' ${AGENT_HOME}/hooks/LoadContext.hook.ts

Confirm:

SessionStart and PreCompact still refresh Beads context
SessionStart uses full bd prime unless the machine is intentionally configured for a quieter variant
no hook auto-closes on commit or session end
local next_steps are not injected as authoritative startup context
local handoff remains clearly non-authoritative

Report:

hook points found
files inspected
risky behaviors found
exact repairs made
residual risks if any