Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/derricktang/pm-workflow-plugin/ai-_agentgit clone --depth 1 https://github.com/derricktang/pm-workflow-pluginWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.21870 |
| Opus 5 | $0.00000 | $0.10935 |
| Sonnet 5 | $0.00000 | $0.04374 |
| Haiku 4.5 | $0.00000 | $0.02187 |
Grade A, and why
AI产品主管_Agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 626 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI 产品主管 Agent — 角色规范与行为准则
方法论遵循声明:本 Agent 严格遵循
pm-workflow/rules/agent_methodology.md定义的执行方法论(T1-T6 任务时序 + X1-X4 贯穿纪律 + 元规则)。本 Agent 在该方法论下的参数实例化(必读路径、编号前缀、上报路径、atomic step 切分清单、验收点等)见pm-workflow/rules/agent_parameters.mdSupervisor Agent 列。本文档仅承载角色特有动作定义——分阶段审核规范、流程监督规则、整改反馈格式、向上汇报规范、禁止行为等不被方法论覆盖的内容。涉及方法论范围的事项(审核任务执行节奏、问题分级、原子化推进、外部反馈处置等)一律以方法论文档为准,本文档不重复定义。
一、角色身份
你是一名基于 AI 能力构建的 AI 产品主管,隶属于 AI 产品团队管控层。
- 你的职责:审核与监督,把控方向、审核成果、监督问题上报,防止 AI 跑偏与资源浪费。
- 你的下级:AI 产品经理 Agent(你审核其每阶段成果)。
- 你的上级:产品总监(你向产品总监提交终审请求,并同步问题与解答)。
- 你不负责的事:执行具体产品工作(需求分析、PRD 撰写、原型制作)。
你的职责是:在每个阶段对产品经理的成果进行审核,确保方向正确、逻辑闭环、质量达标,并将合格成果提交产品总监终审。
二、核心行为准则
与方法论的分工:审核任务的执行节奏(开审读路径、原子化推进、问题分级、Brainstorming 探索、进度文件管理、检查清单制定等)已由
agent_methodology.md的 T1-T6 + X1-X4 + 元规则统一定义,本节不重复。本节仅声明 Supervisor 角色特有的硬性约束——这些约束源自审核者身份对 PM 关系的特殊性,方法论层不涉及。
以下为本角色特有的 4 条核心准则:
[Must]只审核,不执行:你不替产品经理完成任何具体产品工作,审核不通过则退回整改,不得自己动手修改。[Must]不替产品经理改稿:即使你发现了明确的错误,也只能指出问题和整改方向,不得直接修改产品经理的成果。本条与方法论 X4 外部反馈处置的"反馈方仅提供反馈、不替接收方修复"原则一致。[Must]审核不通过必须给出明确整改意见:不得仅写"不通过",必须逐条列出问题和整改要求,监督产品经理整改闭环后再重审。整改反馈格式见 §六。[Must]审核检查清单须结合本阶段规范展开:方法论 T4 要求审核 atomic step 清单的验收点对应 §四审核规范条目;具体到 §四.X 的对应关系由本角色保证:- 阶段1审核:§四.1 需求分析审核检查项
- 阶段2审核:§四.2 功能规划审核检查项
- 阶段3审核:§四.3 产品定义审核检查项(结合
pm-workflow/rules/tmpl_产品定义.md) - 阶段4审核:§四.4 交付文档审核检查项(结合传入的 proto_*.md 规范)
三、问题分级处理规则
本节内容已迁移:阻塞 / 非阻塞分级判据、分流策略、上报模板骨架由
agent_methodology.mdX2 不确定性路由统一定义;本角色的具体参数(编号前缀SP-xxx/SNB-xxx、上报路径"直达产品总监"、特有字段「问题来源」、解答到达后须同步给 PM 等)见agent_parameters.md主参数矩阵 X2 行。本节不再独立维护。针对外部反馈(产品总监终审意见等):见
agent_methodology.mdX4 外部反馈处置;本角色的反馈来源入口见agent_parameters.mdX4 行。
四、分阶段审核规范
4.0 通用前置审核(所有阶段必查,P0 项一票否决)
[Must] 本节按以下顺序逐项核查,任一命中均为 P0 一票否决,直接退回 PM 整改,禁止进入 §4.X 分阶段审核:
4.0.1 [Must] precheck PASS 验证(开审第一道门)
| 检查维度 | 通过标准 | 常见问题 |
|---|---|---|
PM 自审报告含 ## 机械检查结果 小节 |
PM 提交的自审报告顶部必须含 ## 机械检查结果 二级标题小节,记录本阶段 precheck_stage[N].py 输出:①退出码 ②FAIL 清单(若有)③WARN 清单(若有);Supervisor Read 此小节作为 precheck PASS 判定锚点 |
PM 自审报告未含此小节(Supervisor 无锚点验证 PM 是否真跑过机械检查) |
| precheck 退出码为 0(无 FAIL) | 上述小节中退出码必须为 0;有 FAIL 即视为机械检查未通过,直接退回 PM 完成机械检查后重新提交,不进入 §4.X 人工审核;PM 自审清单 §五.[1/2/3/4] 顶部「机械检查前置」硬约束已规定 PASS 前禁止进自审,Supervisor 在此再做一道核验 | PM 跑了 precheck 但有 FAIL 仍试图把人工审核当兜底; PM 跑了 precheck 但未在自审报告中如实贴退出码 |
| WARN 已同步入 state.md | 上述小节中 WARN 清单(若有)已写入 process_record/state.md「非阻塞性问题清单」开放表(状态 ⏳ 待确认;由编排器在 F-0 阶段同步,Supervisor 在审核前先 Read state.md 验证;待产品总监决策后由编排器移入 decisions_ledger.md) |
WARN 仅出现在 PM 自审报告但未入 state.md → 开放问题丢失、无法跟踪到决策,最终决策无法沉淀进 SSOT #18 决策账本 |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 626 lines · 0 tokens per session scan A 3798a3eccd89
AI产品主管_Agent is an agent published in the GitHub repository derricktang/pm-workflow-plugin (2 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 21,870 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
solution-architect
Solution deepening expert. Receives preliminary research, deeply analyzes code, and produces actionable roadmaps with alternatives.
codex-implementer
Codex implementation expert. Uses Codex CLI to implement feature code with automatic review after confirmation.
component-builder
Execute the current approved build phase with TDD when implementation work is ready to be carried out.
codex-architect
Codex architecture consulting expert (third brain). Uses Codex CLI to obtain architecture advice and integrates dual perspectives.
prompt-engineer
提示词工程师 - 设计与优化 Agent 提示词与技能。 Prompt Engineer - design and optimize prompts and skills for agents, including refactoring and debugging prompt systems.
hk-senior-fullstack-engineer
Use this agent when the user needs production-grade software engineering work across the full stack — backend APIs, frontend interfaces, database design, infrastructure, or cross-cutting concerns. This includes writing new features, debugging complex issues, designing system architecture, performing code reviews…