Install
$ agentstack add skill-yli769227-jpg-agent-discipline-agent-team ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Work As An Agent Team / Agent Team 协作
> Prime directive: keep the main context lean. A bloated lead thread is how work derails. > 首要原则:精简主上下文。主线程一臃肿,工作就脱轨。
The discipline / 纪律
- Send non-trivial work to a sub-agent team instead of unrolling huge edits in the main conversation.
非平凡任务派子 agent,而不是在主对话里展开大量编辑。
- The main thread does only: plan, accept, smoke-test, communicate.
主线程只做:规划、验收、冒烟测试、沟通。
The core roles / 核心角色
- Engineer — writes code, makes the thing run. / 工程师:写代码,让项目跑起来。
- QA Inspector — finds the cracks; verifies happy path, error path, and boundaries. / 监理:找裂纹,验证正确/错误/边界路径。
- Lead (main thread) — plans, coordinates, accepts. / 主线程:规划、协调、验收。
- Add specialist agents (Research, Security, …) as the task needs. / 可按需追加 Research、Security 等专职 agent。
QA principle / 监理原则
- QA is not optional — every non-trivial task gets reviewed before "done".
监理不是可选项——任何非平凡任务完成后必须验收。
- QA's verdict has veto power. A FAIL must be fixed and re-reviewed.
监理结论有否决权。FAIL 必须修复后重新验收。
When this triggers / 触发时机
- About to start a main-thread task that will edit 3+ files / span multiple modules.
- About to take on a task that reads 5+ files / generates large amounts of code.
- About to declare a non-trivial task done (it must pass QA first).
Sub-task must-include checklist / 派子任务必含清单
When the lead hands work to a sub-agent, the task description MUST spell out all four — this is the single source of truth the other disciplines point back to: 主线程派子任务时,任务描述必须写明这四项——这是其余纪律在"派子任务"场景下的统一出口:
- Verification step — how to confirm it works (test-is-truth). / 验证步骤——怎么确认跑通(test-is-truth)。
- Key-path logging — connect / state change / error / request-response (log-first). / 关键路径日志——连接/状态变化/错误/请求响应(log-first)。
- Build after each file edit (incremental-build). / 每编一个文件就 build(incremental-build)。
- Re-read the latest file before editing — don't trust a cached copy (no-dead-code). / 编辑前重读最新文件——不信缓存副本(no-dead-code)。
Done criterion (verifiable) / 完成判据(可验证)
✅ Non-trivial work ran through sub-agents, and a QA inspector reviewed it (happy + error + boundary paths) and returned PASS before "done" was claimed. ⚠️ QA returned PASS-with-caveats — surface the caveats. ❌ The lead thread did everything itself and/or skipped QA. Route it through the team and review.
Worked examples / 实战反例
Real before/after cases for this discipline live in [EXAMPLES.md](./EXAMPLES.md) — read them before you act. 本纪律的真实 before/after 反例见 [EXAMPLES.md](./EXAMPLES.md) —— 动手前先对照。
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: yli769227-jpg
- Source: yli769227-jpg/agent-discipline
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.