Investigate
PR-grade investigation of an issue or request — deep dive, options, and a ranked recommendation, no implementation. Use to scope a bug or feature before writing code.
Audit Branch
Critically audit the changes on a branch — verdict, design audit, strategic implications, and ranked recommendations. Use before merging or when reviewing a diff.
Parallel Research
Run public-safe Parallel.ai research from the terminal and write a cited Markdown report for dev, architecture, standards, and prior-art questions.
Linear Issues
Decompose findings into well-scoped Linear issue drafts (one issue = one PR), preview them, and ask before filing. Never files without explicit confirmation.
Audit Findings
Cross-audit an existing list of findings or recommendations — pressure-test each, stabilize modified/new ones, surface what was missed. Use after a review produces findings.
Setup Ad
Set up an architecture-decision archive, agent-context index, and hard-rules table in a repo that lacks them, seeded from choices the code already embodies.
Evaluate Source
Evaluate external content (article, repo, tweet, docs) for ideas worth borrowing into your system — verdict per idea, ranked. Exploratory, no commitments.