Full pipeline: evaluation -> spec -> implementation -> review -> QA
日本語の概要は準備中です。原文の説明を表示しています。
QA + bugfix cycle until it passes
インストール方法を見るインストールする前に、エージェントに与えられる指示の中身を確認できます。
Runs a QA validation cycle followed by bugfix until all criteria pass clean. Useful for validating implementations without the full build-feature pipeline.
/qa-cycle
Before invoking the QA agent, run the project verification commands. The specific commands are in the "CLI commands" section of CLAUDE.md:
npm test) — record result and outputnpm run lint) — record result and outputInvoke the QA agent using Task tool with model: "sonnet" (execution tier):
.claude/agents/qa.md to assume the QA roleIf QA reports bugs (including test/lint failures), invoke the Bugfix agent using Task tool with model: "sonnet" (execution tier):
.claude/agents/bugfix.md to assume the Bugfix roleReturn to Step 1 (automated verification) to re-validate after the bugfix. Maximum 3 verification-QA-bugfix cycles to prevent infinite loops.
Present the result:
User: /qa-cycle
QA (sonnet) — Cycle 1: 2 of 5 criteria pass. Bug: form validation missing on email field.
Bugfix (sonnet) — Added email regex validation to UserForm component.
QA (sonnet) — Cycle 2: 5 of 5 criteria pass. 0 bugs.
Result: Approved.
When spawning QA or Bugfix agents via the Task tool, always use subagent_type: "general-purpose". Guild agent role names are NOT valid Claude Code subagent_types.
Example:
Task tool with:
subagent_type: "general-purpose"
model: "sonnet"
prompt: "Read .claude/agents/qa.md and assume that role. Then: [task description]"
The model parameter is resolved from the step's model-tier: qa-validate and bugfix use execution→"sonnet".
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Full pipeline: evaluation -> spec -> implementation -> review -> QA
日本語の概要は準備中です。原文の説明を表示しています。
Convenes multiple agents to debate an important decision
日本語の概要は準備中です。原文の説明を表示しています。
Create a pull request from the current branch with structured summary
日本語の概要は準備中です。原文の説明を表示しています。
Discipline skill — systematic debugging process. Use when encountering any bug, debug issue, test failure, broken function, or unexpected behavior, before proposing fixes.
日本語の概要は準備中です。原文の説明を表示しています。
Enriches CLAUDE.md by exploring the project and specializes agents to the real stack
日本語の概要は準備中です。原文の説明を表示しています。
Incremental re-specialization — re-scans the project and updates only auto-generated zones in CLAUDE.md and agents
日本語の概要は準備中です。原文の説明を表示しています。