auto-task-criticize
Review a plan or GitHub issue for an `auto_task` before execution
SKILL.md.claude/skills/auto_task.criticize/SKILL.md
GitHubGoal
- The user passes a file
<FILE>(e.g.,tasks.md) or a GitHub issue number<GITHUB_ISSUE_NUM> - Read the content carefully. Make sure the problem and the solution are clear and complete
Workflow
Read Context
- Read the plan (file or GitHub issue)
- Analyze the problem and the solution
Create a Plan, If Needed
Identify the repos the change affects. This decides whether a PR spans multiple repos
- Update the
* Repo:checklist in.claude/templates/auto_task.template.md, following.claude/skills/auto_task.rules.mdsection "Multi-Repo Issues, Branches, and PRs" - If more than one repo is checked, each
PR<NUM>block in "Solution" must name its target repo. An unlabeledPR<NUM>in a multi-repo task is an unclear plan (see "Ask for Clarification Before Executing an Unclear Plan" in.claude/skills/auto_task.rules.md) - Do not add or edit a
## Companion PRssection. Once PRs exist,git_create_issue_and_branch.pygenerates it, per.claude/skills/auto_task.rules.md"Multi-Repo Issues, Branches, and PRs"
- Update the
If the task is not perfectly clear, follow
.claude/skills/auto_task.rules.mdsection "Ask for Clarification Before Executing an Unclear Plan"
Output Results
- Follow
.claude/skills/auto_task.rules.md - Write the output as a file in the format
.claude/templates/auto_task.template.md
Conventions
- Follow
.claude/skills/coding.rules.mdwhen writing code - Follow
.claude/skills/testing.rules.mdwhen writing testing code
Verification
- The problem statement is clear and unambiguous
- The proposed solution is clear and unambiguous
- Any needed clarification was requested before the task moves to execution