Quality

Checkpoint gives a concise status update with the hypothesis tested, exact changes, before/after metrics, and the next action. Use it to summarize progress and blockers during iterative work.

Codex Audit sends recent changes to OpenAI Codex CLI for an independent review. Use it after edits or new commits to check for bugs, security issues, and logic errors, then validate the findings against the code and project docs.

Codex Direct File Audit sends specific files to OpenAI Codex CLI for an independent review. Use it when you want bugs, security issues, performance problems, and logic errors checked against the source.

Facts Discover scans a codebase and updates a fact sheet to match what the code actually shows. Use it to discover, audit, or sync facts by tagging them as draft, spec, implemented, or untagged.

Facts Implement turns @spec facts into code and tags them @implemented. Use it when you need to build from a fact sheet, complete unimplemented requirements, or verify work against the spec.

Pair continues a pair-programming session by reviewing a partner’s work, verifying correctness, making edits, and completing the next task when appropriate. It appends the next message to a file and commits the changes.

Self Audit pauses to review recent work from first principles. It separates facts from assumptions, identifies likely errors, and recommends the highest-priority fix or confirms no changes are needed.

Codeck Review opens rendered slides in a browser, checks each page for narrative, content, and visual issues, and applies scoped fixes in slides or CSS. Use it when you need review, QA, inspection, or proofreading of a deck.

Design Review checks HTML files against the melta UI design system and flags violations, false positives, and severity. Use it when you need a DS compliance review and fix suggestions for a specific file.

Security Review audits changed code for vulnerabilities such as injection, auth bugs, data exposure, and unsafe deserialization. Use it when you need a security assessment or OWASP-focused review.

Weak Agent Test runs an adversarial harness that sends weak agents through real docx-cli document tasks and grades the results in Word. It is used to compare editing and authoring behavior, measure tool use and token cost, and produce a prioritized ergonomics report.

Code Review provides structured reviews of code changes for bugs, security issues, performance problems, and best-practice gaps. Use it when reviewing a diff and you need severity-tagged, actionable feedback without rewriting the code.