Setup

Purpose

Setup per the uniform setup contract (docs/PLUGIN-PHILOSOPHY.md "Setup is explicit and repeatable" in the marketplace repository): check inspects and reports, apply resolves. This plugin declares no userConfig, and has two setup concerns:

• the external command-line tools its bundled scripts require, where apply is guidance-and-verify with no write path: it points at platform install instructions and never installs a system package;
• the tracked consumer-project configuration audit-pass reads, the suppression record at .claude/audit-pass.md, the one surface apply may write.

Action routing: no argument or check runs the check; apply runs the check first, then remediation. Both are non-interactive, so never prompt when the action is given.

check (read-only)

The bundled scripts are the single source of truth for what this plugin requires.

Read it first. Probe what it actually does, don't recite this file. Then run each probe via Bash and report a PASS/FAIL/INFO table with one remediation line per FAIL. Do not modify anything.

The runtime scripts and their tools:

• ${CLAUDE_PLUGIN_ROOT}/skills/audit/scripts/check-plugin-drift.sh: jq and curl, plus awk and sort
• ${CLAUDE_PLUGIN_ROOT}/skills/audit/scripts/check-structure.sh and check-hook-coverage.sh: jq; fix-plugin-drift.sh: jq plus sort
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-automation-gaps/scripts/inventory.sh: jq
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-permission-grants/scripts/permission-rule-check.sh: jq plus awk and sort
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-permission-state/scripts/*.sh: jq (required; permission-state.sh exits 2 without it), awk and sort; automode-block-lint.sh and automode-entry-diff.sh add python3 for an optional lane
• ${CLAUDE_PLUGIN_ROOT}/skills/draft-auto-mode-rules/scripts/draft-automode-block.sh: jq plus awk
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-pass/scripts/run-state.sh and ${CLAUDE_PLUGIN_ROOT}/lib/check-retirements.sh: jq (run-state.sh falls back to python3, then to a scan that announces itself)
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-instructions/scripts/instruction-scan.sh: grep only (POSIX; no jq)
• ${CLAUDE_PLUGIN_ROOT}/skills/audit-instructions/scripts/conflict-scan.sh: awk and sort (no jq)

Only conflict-scan.sh and permission-state.sh probe for their tools; the rest call them with no guard, so read each script's actual calls rather than trusting a single script's prerequisite block to speak for the plugin.

1. jq, via command -v jq. FAIL if absent: the JSON-parsing scripts need it (inventory.sh degrades to an empty inventory; the others exit 2 with an install remediation). Missing jq blocks every skill whose scripts parse JSON (audit, audit-automation-gaps, audit-permission-grants, audit-permission-state, draft-auto-mode-rules, and audit-pass's run-state and retirement helpers); audit-instructions scans markdown and is unaffected.
2. curl, via command -v curl. FAIL if absent, but scoped: only the plugin-drift check (check-plugin-drift.sh) uses it and exit 2s without it. The rest of audit and every other skill still run; say so in the remediation line.
3. awk and sort, via command -v awk and command -v sort. FAIL if either is absent, and not scoped to one skill: conflict-scan.sh executes both and exit 2s naming the missing one, while check-plugin-drift.sh (both), permission-rule-check.sh (both), and fix-plugin-drift.sh (sort) reach them with no prerequisite check at all, so audit and audit-permission-grants fail mid-run on a bare command not found rather than on a named prerequisite. Say in the remediation line that this FAIL reaches three skills, not just audit-instructions. Report awk and sort by name rather than as one row, since a minimal shell can carry one and not the other.
4. Bash shell, INFO: the scripts are bash (arrays, [[ ]], process substitution, BASH_SOURCE), run through Claude Code's Bash tool, which is the bash shell on every platform and Git Bash on native Windows. Report the resolved interpreter; FAIL only if no bash is resolvable.
5. Network reachability, INFO only: audit's drift/freshness fetches read raw.githubusercontent.com, and a failed fetch degrades to SKIP rather than a setup failure. Do not fetch here; check performs no network call.

The audit-pass suppression record

audit-pass reads a tracked suppression record layered per the marketplace's config-cascade convention. Read the operative shape from ${CLAUDE_PLUGIN_ROOT}/skills/audit-pass/reference/suppression.md rather than inferring it. That reference ships inside this plugin, so it resolves in an installed cache where a path out to the marketplace's own docs does not. The cross-consumer key contract is the marketplace's published finding-suppression convention, which is not a runtime dependency of this plugin. All layers absent is a valid state (no suppressions), so report INFO, never FAIL, when none exists.

Anchor at the repo root (${CLAUDE_PROJECT_DIR}, else git rev-parse --show-toplevel), never a CWD-relative read, which resolves a different (or missing) file depending on the subdirectory or nested worktree the skill was invoked from, then report one row per layer. The same tracked/ignored question has opposite correct answers per layer, so verify each on its own terms:

• user-global ~/.claude/audit-pass.md: outside the worktree; no git command applies. INFO only.
• team .claude/audit-pass.md: must be tracked, probed as the pair — git check-ignore -v reports no match (a match is FAIL with the pattern) AND git ls-files --error-unmatch exits 0. Untracked while present is a hard STOP: teammates never receive the shared suppressions.
• local overlay .claude/audit-pass.local.md: must be gitignored and never staged. Staged or tracked is a FAIL: a personal deviation can reach team history.

Parse each present layer and report as malformed any entry missing any of its five required keys: check, claim, sites, reason, date. Checking only reason and date would pass an entry that audit-pass itself rejects, so readiness would report green on configuration that cannot suppress anything. Report as malformed too an entry whose stored constituents do not hash to its own key: the constituents are authoritative and the key is derived from them. A malformed entry does not suppress, and a silent partial parse would turn a formatting slip into a lost check.

Report a user-global or overlay entry for an id the team layer does not carry as INFO personal-only, not applied, naming promotion to the team layer as what makes it take effect. The team layer is the only one that enacts a suppression, so a personal-only entry that looks live is a finding the operator believes is accepted and is not.

apply (idempotent)

Run check, then for each FAIL give the platform install instructions from the README Requirements. This skill never installs system packages:

• missing jq: the platform's jq install (for example winget install jqlang.jq, brew install jq, or apt-get install jq); rerun check after.
• missing curl: the platform's curl install. Modern Windows and Git Bash already ship curl. Only the plugin-drift check needs it, so the rest of the plugin works meanwhile.
• missing awk or sort: both ship with every POSIX userland, so absence means the shell environment is minimal rather than that one package is missing. Git Bash and Windows busybox shims are where this shows up. Remediate by installing a full userland rather than the single tool: Git for Windows, which bundles both; the distribution's gawk/mawk and coreutils on Linux; brew install gawk coreutils on macOS. Report the two separately, since a minimal shell can carry one and not the other. Five skills depend on them (audit, audit-permission-grants, audit-permission-state, draft-auto-mode-rules, and audit-instructions), and only conflict-scan.sh and permission-state.sh exit 2 cleanly, so do not offer the others as still working meanwhile.
• no resolvable bash: also not remediable by one package. The scripts use arrays, [[ ]], process substitution, and BASH_SOURCE, so they need a real bash on PATH: Git for Windows on native Windows, the distribution's bash elsewhere. Nothing bundled runs until it resolves, so say that this FAIL blocks the plugin rather than offering a partial workaround.

The network row stays INFO and has no apply entry on purpose: a failed fetch degrades to SKIP by design, so there is nothing to remediate.

After any install, re-run the relevant check probe and report its actual result. Never claim resolved on the install command's exit code alone. Re-running apply once every probe passes changes nothing and reports "already configured".

Then converge the one surface this plugin owns, conservatively:

• No suppression record anywhere: leave it that way and say so. Absent is valid; an empty scaffold is noise. Scaffold the team layer with the documented shape only on an explicit request.
• Team layer present but untracked: report the STOP and the exact git add the operator should run. Never stage on their behalf.
• Overlay present but not ignored: recommend the recursive .claude/**/*.local.* line and leave the .gitignore edit to the consumer.
• Malformed or unrecognized entries: report them and stop. Never rewrite, reorder, or drop an operator's suppression: an entry this skill cannot reconcile is a question for the operator, and a silent rewrite would hide the very finding the entry was suppressing.

Every write names the file and the exact change before making it, and preserves unrelated content.

Gotchas

• Never run a git command against the user-global layer. ~/.claude/audit-pass.md is outside the worktree, so git check-ignore and git status return a meaningless verdict there, or a confidently wrong one when the home directory is itself a git repository.
• Missing config is not a failure. All three suppression layers absent means no suppressions, which is the normal state for a repo that has never suppressed a finding. INFO, never FAIL.
• Recommend the recursive gitignore line, and leave the edit to the consumer. .claude/**/*.local.* covers flat, folder-form, and profiled overlays alike; the narrower .claude/*.local.* silently misses any nested overlay. The consumer's ignore file is their artifact; this skill never writes it.
• An install command's exit code is not verification. After any install, re-run the probe and report its actual result.

What this skill does NOT do

• Run an audit; that is the plugin's audit and drafting skills (/claude-config:audit, audit-automation-gaps, audit-instructions, audit-pass, audit-permission-grants, audit-permission-state, audit-prompting-postures, draft-auto-mode-rules, unhobble).
• Write the plugin cache, Claude Code user settings, or pluginConfigs.
• Install system packages.
• Write the consumer's .gitignore, stage anything, or edit an operator's suppression entries.
• Download anything. check makes no network call; the audit skills' own doc/marketplace fetches are theirs, not setup's.