From claude-apd
Use when you need to verify APD is correctly configured in the current project. Qualitative audit of agents, hooks, CLAUDE.md, pipeline, and guardrails — goes deeper than verify-apd.
npx claudepluginhub zstevovich/claude-apd --plugin claude-apdThis skill is limited to using the following tools:
> Qualitative review of how APD is configured in the project — content quality,
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Checks Next.js compilation errors using a running Turbopack dev server after code edits. Fixes actionable issues before reporting complete. Replaces `next build`.
Qualitative review of how APD is configured in the project — content quality, not just file existence. Pairs with
verify-apd(mechanical checks).
Use when:
/apd-setup — confirm everything is correctverify-apd passes but something "feels off"Skip when:
verify-apd itself is failing — fix those mechanical issues firstverify-apd is faster| verify-apd | /apd-audit |
|---|---|
| Files exist? | Content correct and complete? |
| JSON valid? | Hook if patterns correct? |
| Agents have model? | Agents have correct model/effort/color/maxTurns? |
| Pipeline runs? | Pipeline output matches expected format? |
| Mechanical ✓/✗ | Qualitative review |
bash ${CLAUDE_PLUGIN_ROOT}/plugins/apd/bin/core/verify-apd
If FAIL → fix those first. This skill builds on top of verify-apd, not replaces it.
For each agent in .claude/agents/*.md:
Frontmatter check:
model: — builders must be sonnet, reviewer must be opuseffort: — builders must be xhigh, reviewer must be maxcolor: — should be set (purple/blue/green/cyan for builders, orange for reviewer)maxTurns: — should be set (40 for builders, 30 for reviewers); legacy 20/15 auto-bumped by apd initpermissionMode: — builders bypassPermissions, reviewer planmemory: project — should be setHook check:
if: field must be inside hook objects, NOT at matcher levelif patterns (e.g., Bash(git *) not Bash(APD_ORCHESTRATOR_COMMIT=1 git *))${CLAUDE_PLUGIN_ROOT}/plugins/apd/bin/core/ pathsBody check:
Check that CLAUDE.md has all required sections:
## Stack — technology table## APD — orchestrator role description### Pipeline — enforced pipeline reference### Guardrails — guard script list### Mandatory skills — brainstorm/tdd/debug/finish table### Human gate — approval requirements### Session memory — session-log reference## Anti-patterns — common mistakes## Memory — @.claude/memory/ referencesCheck that CLAUDE.md does NOT contain:
{{PLACEHOLDER}} unreplaced values/apd-init, /github-projects, /miro-dashboard)superpowers:subagent-driven-development references (should use APD pipeline)Read .claude/settings.json and verify:
enabledPlugins.superpowers@claude-plugins-official: falseattribution.commit: "" (empty — no AI signatures)attribution.pr: "" (empty)permissions.allow includes .claude/memory/** (Write and Edit)Read .claude/rules/workflow.md and verify:
apd pipeline commands (not apd-pipeline)bash .claude/bin/apd pipeline status
bash ${CLAUDE_PLUGIN_ROOT}/plugins/apd/bin/core/apd-init --version
Check .claude/memory/:
MEMORY.md — not empty, has project contextstatus.md — has current phasesession-log.md — exists (may be empty for new projects)[fill in] placeholders in the last session-log entry (blocks new tasks)APD Project Audit — {project name}
CRITICAL:
1. [file:line] Description
IMPORTANT:
1. [file:line] Description
CLEAN:
✓ Agents (X builder + 1 reviewer)
✓ CLAUDE.md sections complete
✓ Settings configured
✓ Workflow rules current
✓ Pipeline healthy
✓ Memory files present
Result: X issues (Y critical, Z important)
| Excuse | Reality |
|---|---|
| "verify-apd passes so it's fine" | verify-apd checks structure, not content quality |
| "Agents work, no need to audit" | Wrong model or missing maxTurns wastes time and money |
| "CLAUDE.md looks ok" | Missing sections mean orchestrator skips important rules |
| "I'll fix it when it breaks" | Broken pipeline produces broken code silently |
You're done when:
{{PLACEHOLDER}} values.claude/settings.json has all four required keys (env, attribution, enabledPlugins, hooks)apd pipeline status runs without error/apd-setup to regenerate missing pieces/apd-setup → escalate to user with concrete file:line references