AgentSkillsCN

plan-all

按顺序运行完整的 Evoloop 规划管道——从启动、区域处理、评审、红队测试，再到 PM 故事生成——并实时追踪项目进展

SKILL.md
--- frontmatter
name: plan-all
description: Run the full Evoloop planning pipeline — start, area processing, review, red-team, and PM story generation — in sequence with progress tracking
disable-model-invocation: true
allowed-tools:
  - Read
  - Glob
  - Grep
  - Bash
  - Edit

Full Planning Pipeline Runner

Run all 5 subphases in order. Stop on any failure.

Arguments

Parse $ARGUMENTS: --tool <tool> (default: claude), --max-area-iterations <N> (default: 5), --dry-run.

Workflow

  1. •Preflight: ./scripts/doctor.sh --planning-only --skip-runner-tools. Fail → suggest /doctor-fix, stop.
  2. •Plan Start: ./orchestrator.sh plan start --tool <tool>. Verify .plan/areas.md exists after.
  3. •Area Processing: Parse .plan/areas.md table. For each non-approved/locked area, run ./orchestrator.sh plan area --area <name> --tool <tool>. Re-read status after each; retry up to max iterations. If area doesn't reach approved → report and stop.
  4. •Review: ./orchestrator.sh plan review --tool <tool>
  5. •Red-Team: ./orchestrator.sh plan redteam --tool <tool>
  6. •PM Stories: ./orchestrator.sh plan pm --tool <tool>
  7. •Final Validation: ./scripts/doctor.sh (full). Pass → show counts, suggest ./orchestrator.sh run. Fail → suggest /validate.

Dry Run

Run preflight only, show areas that would be processed, print command sequence, note already-approved areas.

Error Recovery

Show full output, identify failed step (e.g., "Failed at Step 3/6: Area Processing (area: backend)"), suggest manual retry command. Re-running /plan-all skips already-approved areas.