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 instructions/ooooooooooooooooooop/agent-tools/agents-mdgit clone --depth 1 https://github.com/ooooooooooooooooooop/agent-toolsWhat 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 | $0.01344 | $0.01344 |
| Opus 5 | $0.00672 | $0.00672 |
| Sonnet 5 | $0.00269 | $0.00269 |
| Haiku 4.5 | $0.00134 | $0.00134 |
Grade A, and why
agent-tools AGENTS.md 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 — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — 本仓库浏览与改动导航
这份文件给进入仓库的代理/协作者一个最快的心智模型：这里有什么、哪些能碰、哪些绝不能碰、改动会被哪些检查拦截。详细契约见 docs/architecture.md。
一句话
这是一个 Agent 工具的源码与发布控制面（Skill 包 + MCP 发行包 + DSH 插件包 + 质量脚手架），不是运行时工作区。根目录 top-level 目录属于五层之一。
分层目录（改动前先认层）
| 层 | 目录 | 可否改动 | 改动要点 |
|---|---|---|---|
| ① Skill 包 | skills/<name>/（20 个） |
✅ | 必须登记在 skills.json；布局统一 SKILL.md + agents/openai.yaml + examples/*.md |
| ② MCP 包 | mcp/<name>/ |
✅ | 登记在 mcp.json；自持子目录许可证；agent-switchboard 是修改版，非 MIT |
| ③ DSH 插件包 | dsh/<name>/ |
✅ | 本地 DSH 用户级插件：源码 + 可移植 cordis.patch.yml 片段 + README 一起发布；不设注册表，校验器不核对 manifest，靠链接/markdown 检查兜底；发布内容禁止含本机路径 |
| ④ 质量脚手架 | scripts/ tests/ docs/ _template/ .github/workflows/（skill-quality-gate 是 ① 中的 Skill 包，位于 skills/skill-quality-gate/） |
✅ | 改 validate_repo.py 会影响全部上层契约，谨慎 |
| ⑤ 设备运行层 | .taskflow/ .grepai/ .claude/ node_modules/ 等 |
❌ 禁改/禁提交 | 本地运行态，发布门禁拒绝 |
硬约束（改目录结构的红线）
- Skill 统一收在
skills/<name>/下。scripts/validate_repo.py::discover_skill_dirs()识别skills/子目录里含SKILL.md的目录（为兼容也认可根目录平铺的旧包）。新增 Skill 放进skills/并登记到skills.json的path（./skills/<name>）；两处不同步会让--strict判 FAIL。 - 两个注册表必须与目录一致：
skills.json↔skills/*；mcp.json↔mcp/*。校验器逐一核对。 - 发布门禁：
git diff --check、python scripts/validate_repo.py --strict、python skills/skill-quality-gate/scripts/quality_report.py --root . --strict、仓库回归、MCP 回归全绿才可合并。
改动后必须跑的检查
python scripts/validate_repo.py --strict # 结构 + 注册表 + 许可证 + 链接 + 运行时文件边界
python skills/skill-quality-gate/scripts/quality_report.py --root . --strict # Skill 门禁
python -m unittest discover -s tests -v # 仓库回归
python -m unittest discover -s mcp/agent-switchboard/tests -v # MCP 回归
git diff --check # 空白错误
本地缓存与运行态（当前仓库的"大块头"）
mcp/agent-switchboard/_sdk_probe_deps/曾放 Claude Agent SDK 缓存（约 360MB，含打包的 claude.exe）。已移到~/.cache/agent-switchboard-sdk/，claude_sdk_backend.py从那里探测（可用AGENT_BROKER_CLAUDE_SDK_DEPS覆盖）。仓库不应再出现这种大体积缓存目录。- 任何
state.sqlite、.jsonl、.log、用户路径、会话/任务 ID、模型/认证信息都是运行时态，禁止由mcp.json或 CI 校验放行到发布集。
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 · 52 lines · 1,344 tokens per session scan A b2f669a0fce1
agent-tools AGENTS.md is an instructions file published in the GitHub repository ooooooooooooooooooop/agent-tools (2 stars, last pushed 4d ago), licensed MIT. It adds 1,344 tokens to every session, about $0.0067 per session on Opus 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 2026-08-31.
Other instructions, from other repositories
claude-code-karma CLAUDE.md
Instructions for JayantDevkar/claude-code-karma, covering claude.md, project overview, quick start, clone (no submodule init needed — this is a monorepo) and start api (terminal 1).
claude-code-skills CLAUDE.md
Instructions for levnikolaevich/claude-code-skills, a project described as: Standalone engineering skills for Claude Code and Codex: review, audit, optimization, testing, product discovery, architecture, and safe publishing.
claude-code-eyes CLAUDE.md
Instructions for fcavalcantirj/claude-code-eyes, covering claude-code-eyes — notes for claude code, what it is, load-bearing invariants (do not "simplify" these away), frontmatter rules (enforced by the skill-creator validator) and how to test (everything must pass under bash 3.2).
opinionated-claude-skills CLAUDE.md
Claude Code instructions for Pyroxin/opinionated-claude-skills, covering claude.md, critical requirement, repository structure, pre-publication validation and build system.
autorun CLAUDE.md
Claude Code instructions for ahundt/autorun, a project described as: Claude Code & codex plugin + Gemini cli extension to keep ai running, keep data safe, plan better, and keep running until tasks are done.
ralvaskills CLAUDE.md
Claude Code instructions for ralvarezdev/ralvaskills, a project described as: My personal, ever-growing collection of AI skills for OpenCode and Claude Code. Enforces strict clean architecture and professional standards.