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/seokan-jeong/team-shinchan --skill team-shinchan-ultrawork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。
正在显示 SKILL.md
基于 SOC 职业分类
| name | team-shinchan:ultrawork |
| description | Use when you need to complete tasks quickly with parallel agent execution. |
| user-invocable | false |
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
👦 [Shinnosuke] Ultrawork mode -- maximum parallelization!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
If args > 2000 chars: truncate + warn.
If args contains a recognized preset name, resolve agent list from agents/_shared/team-presets.json:
node -e "
const fs = require('fs');
const presets = JSON.parse(fs.readFileSync('${CLAUDE_PLUGIN_ROOT}/agents/_shared/team-presets.json', 'utf8')).presets;
const input = process.argv[1] || '';
const matched = Object.values(presets).find(p => input.toLowerCase().includes(p.name));
if (matched) console.log(JSON.stringify(matched));
" "{args}"
agents array to constrain which domain agents are assigned in Step 3Recognized preset names: fullstack, backend-api, quality, data-pipeline, security-audit
Task(subagent_type="team-shinchan:nene", model=,
prompt=)
Store as {plan_context}. Skip if task is clear and touches 1-2 files.
If {plan_context} was generated, compute Wave execution order using ontology:
node -e "
const { getWaveOrder } = require('${CLAUDE_PLUGIN_ROOT}/src/ontology-engine.js');
const taskList = /* parse task units from plan_context as [{id, files:[...]}] */;
const { waves, warnings } = getWaveOrder(process.cwd(), taskList);
console.log(JSON.stringify({ waves, warnings }, null, 2));
"
waves = [[task1, task2], [task3], ...] — 같은 Wave 내 태스크는 병렬 실행 가능getWaveOrder가 단일 Wave를 반환하거나 온톨로지가 없으면, 모든 태스크를 동시 병렬 실행 (기존 동작 유지)Store result as {wave_plan}. Pass to Step 3 execution.
If invoked standalone (not as Stage 3 of an active /start/autopilot workflow),
run the START → In Progress transition per agents/_shared/linear-sync.md before
executing. When driven inside a workflow, skip — the workflow owns the issue lifecycle.
No-op if no Linear issue / Linear MCP unavailable.
Task(subagent_type="team-shinchan:shinnosuke", model="opus",
prompt="/team-shinchan:ultrawork invoked.
${plan_context ? 'Pre-planned:\n' + plan_context : ''}
Parallel execution:
1. Break into independent units (or follow plan)
2. Assign to agents in parallel (run_in_background=true)
Routing: Analysis→Shiro/Misae/Hiroshi | Execution→Bo/Kazama | Frontend→Aichan | Backend→Buriburi | DevOps→Masao | Verification→Action Kamen
3. Queue sequential tasks, wait for completion
4. Integrate results + Action Kamen verification
Done when: all TODOs complete, features working, tests pass, no errors. Keep working if not met.
User request: ${args}")
STOP HERE. The above Task handles everything.
After all parallel agents complete:
Run: node src/stagnation-detector.js --jsonl .shinchan-docs/work-tracker.jsonl --window 20
If stagnation: true, surface findings in the summary before handing off to Action Kamen:
"Stagnation patterns detected: {pattern names} — {evidence}. Reviewing before AK handoff."
After integration and Action Kamen verification pass, and if this was a standalone
invocation, run the FINISH → In Review transition per agents/_shared/linear-sync.md
for the same issue (no-op if not Linear-based / driven inside a workflow).