菜单
Use when hypotheses have been generated and need to be converted into an executable stage plan. Produces a standard stage plan that /build-app executes unchanged.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/cbemister/claude-code-utils --skill plan-optimization命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Identify and fix accessibility issues for WCAG 2.1 Level AA compliance, including mobile screen reader support (VoiceOver/TalkBack), touch targets, and keyboard navigation.
Autonomously build an app from staged plans created by /launch-app. Reads plans/active/ stage files and builds each stage with verification, commits, and Slack notifications. Designed for unattended execution via build-app-runner.sh.
Polish components to production quality — interactive states, micro-interactions, and final refinement. Run all three passes or target one area.
正在显示 SKILL.md
基于 SOC 职业分类
| name | plan-optimization |
| description | Use when hypotheses have been generated and need to be converted into an executable stage plan. Produces a standard stage plan that /build-app executes unchanged. |
Convert batched hypotheses from /generate-hypotheses into a standard stage plan. The output uses the exact stage-plan.md template format so /build-app executes it without modification.
CRITICAL — NON-INTERACTIVE: Execute all steps without pausing. Never ask questions, request confirmation, or wait for input. Read state, plan, write output, update state, stop.
Read factory/evolution-state.json:
"planning"CYCLE to current_cycle"planning", display error and stopRead the latest hypotheses file:
factory/hypotheses/hyp-{CYCLE}-*.jsonbatched: true hypothesesRead project context:
CLAUDE.md — project name, stack, conventionsimplementation_sketchFor each batched hypothesis, expand the implementation_sketch into full task specifications:
pages_affected and modificationsnew, modify, or extend annotationsTask 1A (data model) ───────────────┐
Task 1B (utility function) ─────────┤── parallel
↓
Task 2A (component using 1A+1B) ──── sequential
Task 2B (integration) ──────────────── sequential (after 2A)
Based on task type, assign agents from the project's team:
| Task Type | Recommended Agent |
|---|---|
| Data models, API routes | backend-architect |
| UI components, pages | frontend-architect |
| Tests | test-engineer |
| Performance-sensitive | performance-analyst |
| Security-sensitive | security-auditor |
If the project doesn't have specific agents configured, use generic assignments (lead, support, review).
Write plans/active/stage-opt-{CYCLE}.md using the optimization plan template format:
---
title: "Optimization Cycle [CYCLE]: [Combined Hypothesis Titles]"
type: optimization
status: pending
created: [YYYY-MM-DD]
started: null
completed: null
parent: "evolution-cycle-[CYCLE]"
cycle: [CYCLE]
hypotheses: ["hyp-CYCLE-001", "hyp-CYCLE-002"]
baseline_score: [current composite score]
predicted_score: [expected composite score after optimization]
---
# Optimization Cycle [CYCLE]: [Combined Title]
## Context
This optimization implements [N] hypotheses from evolution cycle [CYCLE] to improve the product's composite score from [baseline] to [predicted].
**Hypotheses:**
1. **[Title 1]** — [rationale summary] (predicted +[X] composite)
2. **[Title 2]** — [rationale summary] (predicted +[X] composite)
**Current Score:** [baseline]/100
**Predicted Score:** [predicted]/100
---
## Agent Team Assignments
| Role | Agent | Responsibility |
|------|-------|----------------|
| **Lead** | [agent] | Owns [domain] |
| **Support** | [agent] | Verifies [concern] |
| **Review** | code-reviewer | Validates implementation quality |
---
## Dependencies
### Requires
- Evaluation data: `factory/evaluations/eval-[latest].json`
- Hypothesis specs: `factory/hypotheses/hyp-[CYCLE]-[latest].json`
### Produces
- Optimized codebase ready for preview deployment
- Updated components and pages per hypothesis specs
---
## Parallelization
[ASCII diagram from Step 2]
---
## Tasks
### Task [CYCLE].1: [Task Title]
**Hypothesis:** [hyp-ID]
**Agent:** [assigned agent]
**Verification:** Tier [N]
**Objective:** [what this task accomplishes]
**Files:**
| File | Action | Purpose |
|------|--------|---------|
| `src/components/X.tsx` | new | [what it does] |
| `src/app/page.tsx` | modify | [what changes] |
**Specification:**
[Detailed implementation spec — TypeScript interfaces, component structure, behavior]
**Verification Steps:**
- [ ] [Tier-appropriate checks]
---
### Task [CYCLE].2: [Task Title]
[Same structure...]
---
## Rollback Instructions
If this optimization is rejected:
1. Preview branch will be deleted by `/evolution-gate reject`
2. No changes reach production
3. Hypotheses are marked as rejected in evolution state
4. Next `/generate-hypotheses` run will skip similar approaches
---
## Stage Checkpoint
| Criterion | Target | Gate |
|-----------|--------|------|
| All tasks complete | 100% | hard |
| Verification passing | All tiers | hard |
| No new lint/type errors | 0 | hard |
| Predicted score achievable | +[X] composite | soft |
Invoke the /plan-next-stage pattern — re-read the actual source files and verify:
new)factory/evolution-state.json"building"{
"cycle": [CYCLE],
"status": "building",
"evaluation_score": null,
"hypothesis_ids": ["hyp-CYCLE-001", "hyp-CYCLE-002"],
"hypothesis_titles": ["Title 1", "Title 2"],
"stage_plan_ref": "plans/active/stage-opt-[CYCLE].md",
"preview_branch": null,
"score_delta": null,
"rejection_reason": null
}
================================================================
Optimization Plan Created
================================================================
Project: [name]
Cycle: [CYCLE]
Hypotheses: [count] batched
Tasks: [total tasks]
Parallel: [count parallelizable] | Sequential: [count sequential]
Plan: plans/active/stage-opt-[CYCLE].md
Baseline: [X]/100
Predicted: [Y]/100 (+[delta])
Next: Run /build-app to execute the optimization plan
================================================================
| Scenario | Recovery |
|---|---|
| No hypotheses files | Display error: "Run /generate-hypotheses first" |
| Source files moved/renamed | Update paths in plan, add note about drift |
| Hypothesis references deleted code | Adjust task to create from scratch instead of modify |
| plans/active/ doesn't exist | Create it: mkdir -p plans/active |
| Context exhausted mid-planning | Runner re-invokes; check for partial plan, resume |