| name | audit |
| description | Structured audit of the Parallax codebase. Run periodically or before major milestones. |
| disable-model-invocation | true |
/audit -- Project Audit
Structured audit of the Parallax codebase.
When to Use
Invoke /audit periodically or before major milestones to check project health.
Checks
Perform each check and report findings:
- Code review: Bugs, concerns, anti-patterns, security issues in
src/parallax/.
- Test gaps: Untested features, weak assertions, missing edge cases in
tests/.
- Doc staleness: Do README.md, CLAUDE.md, ROADMAP.md, VISION.md match current code? Are references valid? For deep doc auditing, run
/doc-sync.
- Dependency health: Outdated or vulnerable dependencies in pyproject.toml/pixi.toml.
- Architecture: Does code match the 3-layer design in VISION.md? Any drift?
- Scientific rigor: Are hypotheses documented? Results reproducible? Assumptions stated?
Parallax-Specific Checks
- ROADMAP.md: are completed items checked off? Are open questions still relevant?
- README.md: does "Current Status" section reflect reality?
pixi run check passes cleanly (lint, typecheck, test).
- Session summaries in
docs/sessions/ are up to date.
- No TODO markers in production code or skills.
Output Format
## Audit: [date]
### Critical
- [findings that need immediate attention]
### Warning
- [findings that should be addressed soon]
### Info
- [observations and suggestions]
### Summary
[1-2 sentence overall assessment]
Rules
- Check @CONSTITUTION.md principles against actual practice.
- Flag any tests that appear to have relaxed tolerances.
- Verify reproducibility claims (are environments locked? versions pinned?).
- Be specific: include file paths, line numbers, concrete evidence.
- Run
pixi run check as part of the audit.