oopsla5xx

Review PR diff against the spec before shipping — check spec compliance, conventions, scope, required security checklist, and test coverage. Outputs a structured verdict (APPROVE / REQUEST CHANGES / COMMENT). Run after self-check, before Phase 4 ship.

Systematic bug investigation workflow. Tool-agnostic but MCP-aware — test cases and MCP tools are evidence sources, not prerequisites. Applies to any stack (frontend, backend, CLI, mobile, distributed systems).

Decide whether to ask before coding or proceed with stated assumptions. Small/clear tasks → do it. Large/ambiguous tasks → state assumptions and ask at most 1 critical question. Run at the start of any task.

Generate a pull request description by filling in the .github/PULL_REQUEST_TEMPLATE.md from spec, git diff, and commit history. Outputs ready-to-use text or creates PR via gh CLI. Run as part of Phase 4 ship.

Implement each step from the plan using TDD — write failing test first, then minimal code to pass, then refactor. Run for each unchecked step in the Plan section of .ai/tasks/<name>.md.

Break a spec file into ordered, concrete implementation steps with done criteria. Appends a Plan section to the existing .ai/tasks/<name>.md. Run after write-spec, before coding.

Bootstrap .ai/ context files — auto-detect stack and fill commands.md, architecture.md, conventions.md instead of filling TODO placeholders manually. Run when setting up oops-skills template in a new project.

Final ship step — push branch, open PR on GitHub, update status.md, clean up task files. Run after review-pr APPROVE and create-pr-description.

Scan the project and generate a context.yaml draft — reads oc-settings.yaml if present, otherwise infers everything from docs and source code. Runs validate automatically after generation.

First-run setup wizard for open-context — asks 2 questions (scope, communication language), then reads the project's own docs (or source code, if none exist) to synthesize a project profile (stack, architecture, actors) for a single confirm — then generates…

Manually resolve a task against the project's context.yaml — shows which domains, components, files, and rules would be injected by the hook. Use for debugging routing and keyword coverage.

Run phrasing coverage tests and amplification safety checks against the project's context.yaml. Reports per-domain pass/fail rates, risk levels, and amplification warnings.