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 skills/wu736139669/agent-taskgraph-protocol/agent-taskgraphnpx skills add wu736139669/agent-taskgraph-protocol --skill agent-taskgraphgit clone --depth 1 https://github.com/wu736139669/agent-taskgraph-protocolWhat 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 | $0.00126 | $0.03195 |
| Opus 5 | $0.00063 | $0.01597 |
| Sonnet 5 | $0.00025 | $0.00639 |
| Haiku 4.5 | $0.00013 | $0.00319 |
Grade A, and why
agent-taskgraph 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.
This is a copy
100% identical to agent-taskgraph — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 208 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent TaskGraph
只面向 Codex 和 Claude Code。优先使用当前宿主的原生多 Agent 能力，不要求 HAPI。
1. 默认原则
- 一句话可以开始。先读项目，再问代码无法回答的问题。
- 简单任务由当前会话直接完成，不成立团队。
- 复杂或可并行任务才成立团队；主会话成为 PMO，不写产品实现。
- 一个 Agent = 一个独立 Agent thread/session + 一个明确职责 + 一个当前任务。
- 每个 Agent 只加载自己的最小上下文，不复制 PMO 全部聊天历史。
- Agent 之间通过稳定共享文档交接，不以互相转发长聊天作为事实源。
- 默认使用当前 Codex/Claude Code 的原生模型、权限、会话和等待机制。
- HAPI 默认关闭；只有用户明确说“使用 HAPI”或确有跨机器控制需求时才加载
references/hapi-runtime.md。
2. 用户看到的流程
PMO 先输出一段短摘要：
我的理解：<目标与当前项目事实>
需要你决定：<0-3 个真正阻塞的问题，附推荐>
建议团队：<当前会话直接做 / PMO + N 个 Agent>
需要团队时，再展示一次团队预览：
| Agent | 长期职责 | 本次任务 | 原生运行形态 | 模型/推理 | 权限 | Worktree/写入范围 |
|---|
同时用一句话说明 PMO 只做计划、派发、监督和验收。用户确认“开始”后一次性创建团队；不要让用户确认 Goal 路径、Session ID、wait primitive、runtime evidence 或内部命令。只有运行形态、可见方式或权限扩大时才重新确认。
最短用法：
使用 agent-taskgraph。先理解这个项目和需求；如果值得并行，就由你作为 PMO 创建几个原生 Codex/Claude Agent 会话，分别完成、交接和验收；如果不值得并行，就当前会话直接完成。先告诉我理解和分工，确认后开始。
3. 什么时候成立团队
使用当前会话直接完成：
- 局部 bug、单文件或同一模块小改动
- 顺序工作，拆开后没有速度收益
- 多人会同时修改相同文件
- 任务本身还没聊清楚
使用 PMO + 原生 Agent 团队：
- 至少两个任务能独立推进和验收
- 大范围只读探索、测试、审查或资料分析可并行
- 不同模块有清楚且不重叠的写入范围
- 用户明确要求多 Agent、多会话或团队协作
默认 2-3 个工作 Agent，最多按宿主和项目能力扩展。更多 Agent 不等于更快；没有当前任务的角色不创建。
高风险的迁移、权限、支付、删除、发布或共享基础设施变更，在普通团队流程上增加完整规格、独立 Reviewer 和 Human Gate。不要让所有日常任务承担这套成本。
4. 原生运行时
Codex
- 优先使用当前会话暴露的原生 subagent/agent-thread 能力。
- CLI 中 Agent thread 可由
/agent查看；App/IDE 使用其原生 Agent 面板。 - Codex 自己负责 spawn、wait、steer、stop 和结果回传；Skill 不再用 HAPI 模拟这些能力。
- 原生 Agent thread 默认不是新 Terminal，也不保证自动隔离 checkout。读任务可共享；并行写任务必须使用宿主 worktree 或互不重叠的路径。
- 自定义长期岗位可放在
.codex/agents/，但只在角色确实反复使用时创建。
Claude Code
- 复杂团队任务优先使用 Agent Teams：主会话是 Lead/PMO，teammates 拥有独立 Claude Code 会话和独立上下文。只在已启用实验能力时使用；未启用时说明降级，不静默把“多会话团队”替换成 subagents。
- 需要跨 PMO 恢复、长期后台运行或独立 worktree 的 Worker，优先使用 Agent View background sessions；
claude agents是管理界面，Worker 不需要绑定单独终端。 - Subagents 只用于一次性探索、测试、日志分析或 Reviewer 等短支线；它们不算长期 Team Worker。
- 默认显示是当前终端内的 Agent 面板或后台列表。只有用户明确要求可见分屏时才使用 tmux/iTerm2；不要自动打开多个 Terminal.app 窗口。
- Agent Teams teammates 不自动隔离 worktree；并行写任务必须分区。需要强隔离时改用 background sessions 或支持 worktree isolation 的 subagents。
What ships with it
36 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 339 B
- init.sh 3.8 KB runs code
- references/dispatch-bootstrap.md 3.2 KB
- references/hapi-runtime.md 11 KB
- references/native-runtimes.md 6.4 KB
- references/runtime-profiles.md 5.9 KB
- scripts/check-operational-health.py 11 KB runs code
- scripts/check-update.sh 3.2 KB runs code
- scripts/hapi-hub-session.py 31 KB runs code
- scripts/open-worker-terminal.sh 8.3 KB runs code
- scripts/prepare-task.py 18 KB runs code
- scripts/render-dispatch.py 9.0 KB runs code
- scripts/validate-graph.py 7.6 KB runs code
- scripts/validate-state.py 54 KB runs code
- scripts/verify-hapi-session.py 7.0 KB runs code
- templates/context.md 1.4 KB
- templates/DECISIONS.md 411 B
- templates/dispatch.md 2.0 KB
- templates/goal.md 4.2 KB
- templates/graph.yaml 1.4 KB
- templates/ledger.md 3.4 KB
- templates/PLAN.md 406 B
- templates/PROJECT.md 1.5 KB
- templates/report.md 861 B
- templates/role.md 2.2 KB
- templates/ROLES.md 798 B
- templates/spec.md 1.7 KB
- templates/staffing-change.md 2.3 KB
- templates/STATUS.md 532 B
- templates/task-manifest.json 1.7 KB
- templates/task.md 955 B
- templates/TEAM.md 420 B
- VERSION 14 B
- workers/log-cleanup.sh 2.3 KB runs code
- workers/parse-worker-log.py 2.8 KB runs code
- workers/watch-worker.sh 1.2 KB runs code
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 · 208 lines · 126 tokens per session scan A 12307db59cf8
agent-taskgraph is a skill published in the GitHub repository wu736139669/agent-taskgraph-protocol (4 stars, last pushed 19d ago), licensed Apache-2.0. It adds 126 tokens to every session and 3,195 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to agent-taskgraph, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
fabric-exec
Troubleshooting and advanced API reference for fabricexec TypeScript programs, dynamic providers, agents, and schema recovery. Routine pi. coding calls are documented by ambient guidance; load this skill only after an argument-shape error or when an advanced surface needs exact contracts.
fabric-fusion
Multi-model deliberation. Two to 8 distinct models answer in parallel with web-capable tools, then a judge compares consensus, contradictions, coverage gaps, unique insights, and blind spots. Act mode runs 1–4 read-only references, then one actor reconciles and executes. Use when the cost of being wrong justifies…
fabric-rlm
Recursively decomposes oversized tasks into bounded child Pi agents with fresh context windows. Use for whole-repo audits, massive-context analysis, and multi-file refactors that do not fit one context.
fabric-schema
Uses Fabric's typed Schema evidence loop and, when enabled, its bounded local-file transaction channel. Use when surprise must void a plan and mutation claims need explicit postconditions.
fabric-workflow
Runs a dynamic Pi Fabric workflow with code-held phases, fan-out, pipelines, structured agents, and best-effort verification. Use for large audits, migrations, parallel research, or explicit workflow requests.
fabric-council
Runs a bounded multi-perspective Pi Fabric council with independent reviewers and best-effort synthesis. Use for architecture choices, plans, reviews, and adversarial cross-checking.