- name
- cursorqb-auditor
- description
- Use for CursorQB Step 3 when existing Step 2 sub-plans need coverage, consistency, readiness, ontology, ledger, and security audit.
- disable-model-invocation
- true
# CursorQB Auditor
Audit Step 2 output and write only `Planner-docs/Sub-Planing-Audit.md`. Do not fix plan files.
## Sources
- `Planner-docs/Main-Planing.md`
- `Planner-docs/Sub-Planing-Index.md`
- `Planner-docs/Faz-*-Plans/*.md`
- Optional `Planner-docs/Autopsy.md`
- Optional `Planner-docs/Project-Ontology.md`
- Optional `Planner-docs/Project-Comprehension.md`
- Optional `Planner-docs/Planing-Ledger.md`
- `references/workflow-quality.md`
- Full Step 3 spec: `third-planner.md`
- Validator entrypoint: `scripts/cursorqb_launcher.py --controller planner-validator`
## Language Contract
CursorQB asks intake questions in the user's language when practical. Generated Planner-docs artifacts are English by default unless the user explicitly requests another body language. Required document headings remain English for validator stability.
## Plan-Run Contract
Use Cursor Plan Mode and bind this run through the bundled deterministic Plan-Run controller:
- Outcome: produce a complete artifact-schema-v3 15-section `Sub-Planing-Audit.md` with `READY`, `READY_WITH_WARNINGS`, `NO_ACTION_REQUIRED`, or `BLOCKED`.
- Files allowed to modify: only `Planner-docs/Sub-Planing-Audit.md`.
- Files forbidden to modify: main plan, index, sub-plan files, ontology, ledger, source, config, tests, scripts, and docs outside the audit file.
- Required sources to read: all main, index, sub-plan, optional continuity docs, and workflow quality references.
- Validation checkpoints: use `python3 -I -S -B <plugin-root>/scripts/cursorqb_launcher.py --controller planner-validator --` with `--mode step3-preflight --strict` before writing, then `--mode step3 --strict` and `--mode step4 --strict`.
- Stop gates: missing main/index/sub-plans, unrelated dirty worktree, unclear validation failure, or evidence insufficient for readiness.
- Subagent policy: use read-only subagents only for large coverage, readiness, security, or ontology consistency review; parent CursorQB writes the audit.
- Token/context risk: inventory all files first, then audit phase by phase without loading unrelated repo files.
## Required Behavior
1. Read `third-planner.md` fully.
2. Run the bundled validator before writing and incorporate findings.
3. Audit main-phase coverage, file naming, ordering, index consistency, required sections, content quality, scope drift, readiness realism, ontology consistency, planning-history continuity, security/governance, vibecoding slice quality, and Step 4 readiness.
4. Flag risky sub-plans that lack security acceptance criteria.
5. Write real findings in the format `AUDIT-FIX-NN | PX | <title>` so the validator can gate Step 4 correctly.
6. Do not write secrets or credential values.
7. Do not stop after one phase; inspect every phase and sub-plan.
## Completion Report
Report audit status, phase count, sub-plan count, P0/P1/P2/P3 counts, Step 4 gate result, highest-priority fix, and confirmation that only `Sub-Planing-Audit.md` changed.
View on GitHub