| name | impl |
| description | Implement changes from a provided Markdown issue/requirements file, then review the implementation for major issues and iterate until the review passes. Use when the user wants a workflow that takes an .md file path as input, applies code changes based on the file contents, and performs a self-review loop. |
Impl
Overview
Implement based on a Markdown file (issue/requirements) and run a self-review loop that blocks on major issues until the review is clean.
Workflow
Read the Markdown input
- Open the file path the user provides.
- Extract: problem statement, expected vs actual behavior, repro steps, acceptance criteria, constraints, and affected components.
- If critical details are missing, ask focused questions before coding.
Interview for missing requirements (if needed)
- Ask only what is necessary to proceed safely.
- Prefer short, targeted questions; avoid long questionnaires.
- Example questions:
- "What is the expected behavior in this edge case?"
- "Which environment/versions should this support?"
- "Are there constraints on dependencies or configs?"
- "Is there an acceptance test or repro script to use?"
Implement the change
- Identify the minimal set of files to touch.
- Apply changes that directly satisfy the Markdown requirements.
- Prefer small, safe diffs; add tests if the change warrants it.
- Keep edits aligned with the repo's conventions.
Run a review pass (major issues only)
- Check for correctness, edge cases, regressions, and missing requirements.
- Validate error handling and failure modes.
- Flag missing tests only if they create real risk.
- Do not nitpick style; focus on impactful issues.
Iterate until review passes
- If review finds major issues, implement fixes and re-review.
- Repeat until the review has no major findings.
- Summarize the final changes and any remaining risks.
Review Checklist (Major Issues)
- Requirement mismatch or missing acceptance criteria
- Incorrect control flow or logic errors
- Broken error handling, resource leaks, or unsafe behavior
- Incomplete integration (e.g., feature added but not wired up)
- Tests needed to prevent regressions in critical paths
Output Expectations
- Provide a concise change summary.
- List any tests run or suggest tests if none were run.
- If clarifications are needed, ask targeted questions.