AI Fundamentals is a learning resource about the foundations and infrastructure behind artificial intelligence, including GPU architecture, CUDA programming, large language models, AI systems, and related operations. It is intended for AI engineers, system architects, GPU developers, application developers, and researchers. The catalogue add-ons support learning or working with these AI infrastructure topics.
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 skills add ForceInjection/AI-fundamentals --skill multi-session-collabgit clone --depth 1 https://github.com/ForceInjection/AI-fundamentalsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/forceinjection/ai-fundamentals/multi-session-collab)<a href="https://agentmods.dev/skills/forceinjection/ai-fundamentals/multi-session-collab"><img src="https://agentmods.dev/badge/skills/forceinjection/ai-fundamentals/multi-session-collab/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/forceinjection/ai-fundamentals/multi-session-collab"><img src="https://agentmods.dev/badge/skills/forceinjection/ai-fundamentals/multi-session-collab.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00145 | $0.01296 |
| Opus 5.5 | $0.00058 | $0.00518 |
| Sonnet 5.5 | $0.00029 | $0.00259 |
| Haiku 4.5 | $0.00015 | $0.00130 |
Grade A, and why
multi-session-collab 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 today.
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 — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
多会话协作（评审 / 修改 / 测试 / 裁决）
一句话：把「几个人肉传话的 Claude 窗口」组织成有角色、有协议、有裁决的多会话流程。
何时用 / 不用
用：
- 用户要开多个 CLI 窗口分工（评审、修改、测试），并要一个裁决者
- 需要跨会话消息协议、judge rubric、质量门禁
- 要把「评审 → 修改 → 测试 → 裁决」固化成可重复流程
不用：
- 单会话内的并行子任务 → 用后台 subagent（Agent 工具）
- 一次性短任务 → 多开会话只会把协调成本翻倍
- 只是远程盯屏 → Remote Control / Desktop
先问三个问题（选路线）
- 要保留多个窗口？→ 跨会话消息（原生，零配置）
- 接受合并到一个窗口？→ Agent Teams（实验，需
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1） - 同一流程反复跑？→ Dynamic Workflow（脚本化，对抗式核验）
产品化给别人用，再看 Agent SDK。MCP 不是会话通信方案（工具协议，无 agent-to-agent 能力）；除非需要跨重启持久化任务板或非 Claude 客户端接入，不要自建 MCP 总线。
操作流程
A. 跨会话消息（多窗口）
- 命名四个会话：
claude --name reviewer/modifier/tester/judge（或会话内/rename）；也可以claude --bg起后台会话当 worker（见 setup.md） - 定入站策略：按 references/setup.md 的权限矩阵设置
crossSessionInbound；无人值守的-pworker 必须设accept，否则消息进不来 - 下发任务卡、收回回报：模板与 findings schema 见 references/protocol.md
- judge 按 rubric 出三档裁决：通过 / 打回（附整改单）/ 升级人工
- 排错查 setup.md 的故障表；启动命令与设置清单：
bash scripts/print-setup.sh - 纪律：不轮询（用
notify_when_idle）；消息是纯文本，文件要写路径；@引用在接收侧只是字面文本
B. Agent Teams（单窗口编队，实验）
- 设
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1；副作用：带name启动的 subagent 会自动变成 teammate - 按角色卡组队：reviewer 只读 / modifier 可写 / tester 执行 / lead 即 judge（禁止 judge 改代码）
- 共享任务列表 + mailbox；teammate 完成自动把最终答案发给 lead
- 门禁：
TeammateIdle/TaskCompletedhooks（退出码 2 阻断） - 限制：仅交互模式、无 worktree 隔离（按文件/目录分区）、无 resume、官方建议 3–5 人
C. Dynamic Workflow（可重复流水线）
- 用
agent()/parallel()/pipeline()编排，agent()带schema输出 JSON - 评审 → 逐条对抗核验 → judge 合并，只保留经得起反驳的 findings
- 限制：运行中无人工介入、脚本无文件系统/shell、默认 16 并发；可保存为
/命令复用
铁律（安全边界）
- 跨会话消息不能代批权限、不能改配置或 CLAUDE.md、斜杠命令不执行
- judge 不改代码（独立性）；reviewer 只读；角色按文件分区防覆盖
- 每个 agent 独立计费；teams / workflows 在 CLI 里没有默认硬性预算上限，人数与阶段数要克制
- 裁决必须可检查：打回附「文件行 + 验收标准」，否则下一轮修改没有终点
快速参考
- 任务卡 / 回报模板 / findings schema / rubric：references/protocol.md
- 版本要求、命名、权限矩阵、故障排查：references/setup.md
- 启动清单脚本：
bash scripts/print-setup.sh - 背景读物（手册全文）：https://github.com/ForceInjection/AI-fundamentals/blob/main/08_agentic_system/multi_agent/docs/claude-code-multi-session-playbook.md
- 设计说明与实测记录：https://github.com/ForceInjection/AI-fundamentals/blob/main/08_agentic_system/agent_skills/docs/multi-role-agent-skill.md
What ships with it
4 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.
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.
- today Changed c739099c193d
- yesterday First seen · 71 lines · 145 tokens per session scan A f91f7c6923a1
multi-session-collab is a skill published in the GitHub repository ForceInjection/AI-fundamentals (3,083 stars, last pushed yesterday), licensed Apache-2.0. It adds 145 tokens to every session and 1,296 once invoked, about $0.0006 per session on Opus 5.5. 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 2026-10-19.
Other skills, from other repositories
story-readiness
Checks whether a story is ready to implement: clear acceptance criteria, no open questions, valid ADR and dependencies. Use when the user asks if a story, or the sprint's stories, are ready to start or build.
recipe-create-meet-space
Create a Google Meet meeting space and share the join link.
workthreads
SpecStory Workthreads - a weekly work-thread rollup across a team's repos from SpecStory coding histories (any agent - Claude Code, Codex, Cursor, Gemini, and more). It groups the window's sessions into threads of work per project and labels each new / open / recently closed, so a lead sees what shipped, what is still…
atmos-config
Atmos root configuration: atmos.yaml discovery, precedence, deep merging, basepath, imports, minimal bootstrap, and routing to narrower Atmos skills.
handle-linkedin-connection-request-signal
Use this skill when someone sends a team member an inbound LinkedIn connection request. An inbound request is a deliberate, ACTIVE first-party intent signal — meaningfully stronger than a passive profile view — and it deserves different scoring rules. The skill resolves and enriches the requester, gates them through…
remove
Remove a deployed framework or addon from the current workspace.