メニュー
Review the AGENTS.md file to ensure all rules are up-to-date, removing any redundant or overly vague rules and updating deterministic guardrails as needed.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Fully implement a PRD's issues end-to-end — a parallel read-only investigation pass, then strictly serial TDD implementation on the PRD's shared feature branch, HITL issues left for a human. Use after write-a-prd + prd-to-issues, when the user wants to "implement the PRD", run the whole issue queue, or run /implementation.
Check NuGet analyzer packages for newer versions and, if any are found, autonomously drive the entire release end-to-end — PRD and issues, /implementation, cleanup, changelog, PR with auto-merge, tag, and a published NuGet release. Use when the user wants to bump analyzer dependencies or asks "are there any NuGet updates?".
Pick the next AFK issue from issues/ and implement it end-to-end via TDD, then commit. Use when the user wants to advance the ralph queue, asks to "work on the next issue", or runs this in a /loop.
Implement a specific piece of work described in the skill argument, end-to-end via TDD, then commit. Use when the user wants to work on a described task rather than picking from the issue queue.
Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.
Triggered when the user is skeptical or suspicious of a previous response. The
| name | review-agents-md |
| description | Review the AGENTS.md file to ensure all rules are up-to-date, removing any redundant or overly vague rules and updating deterministic guardrails as needed. |
Please review the current AGENTS.md file. Remove any redundant or overly vague rules that are no longer necessary or that the latest model versions handle automatically.
Specifically ensure that any deterministic guardrails are updated to ensure they continue to be effective in guiding the model's behavior.
If a conversation has just completed successfully, review our agreed outcomes. If a new constraint, preference, or architectural decision was made, reflect on it and propose an update to the rules in agents.md.
Always wait for my approval before saving.
Put commands early: Put relevant executable commands in an early section: dotnet build, dotnet test, dotnet format. Include flags and options, not just tool names (e.g. dotnet test --filter, dotnet build -c Release). Your agent will reference these often.
Code examples over explanations: One real code snippet showing your style beats three paragraphs describing it. Show what good output looks like.
Set clear boundaries: Tell AI what it should never touch (e.g., secrets, vendor directories, production configs, or specific folders). “Never commit secrets” was the most common helpful constraint.
Be specific about your stack: Say “.NET 8 with ASP.NET Core, EF Core, and xUnit” not “a .NET project.” Include the target framework and key dependencies.
Cover six core areas: Hitting these areas puts you in the top tier: commands, testing, project structure, code style, git workflow, and boundaries.
Aim for a maximum of 150 lines of instructions: Long enough to be specific, but short enough to be read and followed. If you need more, consider using progressive disclosure and link to separate files for detailed guidelines thereby keeping AGENTS.md focused on the most critical rules and examples.
Do not reference in flight work in the current issues/PRD.md or issues/NNN-some-issue.md files. AGENTS.md should contain timeless rules and examples that are not tied to specific in-flight work. Avoid mentioning specific issue numbers, PRDs, or temporary architectural decisions that may change frequently. Focus on general principles and practices that will remain relevant over time.
