autoreview

Test code, docs, claims, or visuals against an explicit acceptance contract using primary evidence and task-based adversarial passes.

Create a curated, shareable review bundle from a request and relevant repository files.

Create or refine publication-quality figures in matplotlib or TikZ for papers, technical reports, and professional presentations.

Teach a topic through focused lessons and practice, or continue an existing teaching workspace.

Audit an active goal or research branch against its plan, todo, git history, results, and reproducible evidence.

Stress-test a plan, decision, or idea through an interview when the user asks to be grilled or challenged.

Write a handoff document when the user wants another agent or session to continue the work.

Create reusable feature-goal planning scaffolds for Codex goal sessions.