ワンクリックで
Implement PRD stories until all pass, with verification and progress tracking.
npx skills add https://github.com/japurcell/skills --skill prd-buildこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストール
Implement PRD stories until all pass, with verification and progress tracking.
npx skills add https://github.com/japurcell/skills --skill prd-buildこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストール
Implement failing PRD stories with required simplification, review, verification, and progress tracking.
Implement one failing PRD user story, run checks, update progress, and commit.
Review a pull request, branch, recent commits, staged or unstaged changes, or AI-generated code for correctness, readability, architecture, security, performance, standards, spec adherence, and maintainability. Use when reviewing a PR before merge, reviewing changes since a fixed point, reviewing work-in-progress changes, reviewing another agent's code, asking to "review since X", requesting a thermo-nuclear / thermonuclear / deep code quality / especially harsh maintainability review, or when machine-readable output is needed.
Create a Product Requirements Document (PRD) for a feature. Use for requests like create a PRD, write a PRD, plan this feature, requirements for, spec out.
Run the pre-launch checklist via parallel fan-out to specialist personas, then synthesize a go/no-go decision
Chooses the least powerful subagent model that can reliably finish delegated work. Use whenever you are about to launch, spawn, delegate to, or set `model:` for a subagent, task agent, background agent, or parallel worker. Use this before any task-tool launch that needs model selection, especially for exploration, test execution, grading, code review, debugging, or implementation research where a fast cheap model often suffices.
| name | prd-build |
| description | Implement PRD stories until all pass, with verification and progress tracking. |
You are the orchestrator for a software project.
Continue until every story in prd_file has passes: true.
prd_file is the only source of truth for story status.progress_file is supplementary only.<promise>COMPLETE</promise>Orchestrator
passes: falseprd_file, progress_file, and nearby AGENTS.mdImplementer
implementer.implementer; never fix it directly.passes: true in prd_file, orprd_file (required)progress_file (optional): default dirname(prd_file) + "/progress.txt"subagent-model-router skill.progress_file.## Codebase Patterns.## Codebase Patterns at the top.For the highest-priority story in prd_file with passes: false:
Implement
implementer with ./implementer-prompt.mdprogress_file, and mode: implementationVerify
implementerprogress_file, mode: verification_fix, and full failure detailsImprove process
self-improve once, if not already invokedRecord Perform in this order:
AGENTS.md with reusable guidance only.prd_file to passes: trueprd_filepasses: truepassespasses: true, append a progress entry to progress_file.prd_file.passes: false, continue with the next highest-priority story.Mark a story complete only if all are true for the current state:
implementerprd_file are satisfiedImplementer DONE, confidence, or progress_file updates do not satisfy this gate.
Use checks required by:
AGENTS.md## Codebase Patterns at the topRequired entry format:
## [Date/Time] - [Story ID]
- What was implemented
- Files changed
- **Learnings for future iterations:**
- Patterns discovered
- Gotchas encountered
- Useful context
---
Add only reusable guidance, such as:
Do not add story-specific notes.
Stop only if:
prd_file have passes: true, orprd_file has contradictions, invalid ordering, or missing required details that need human correctionBefore any response other than <promise>COMPLETE</promise>:
prd_filepasses: false and no Stop Condition applies, continue the looppasses: falseimplementerprd_file as the official completion sourcepasses: true