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/ctrixin/agent-2-agent/claude-mdgit clone --depth 1 https://github.com/CtriXin/agent-2-agentWrote 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/instructions/ctrixin/agent-2-agent/claude-md)<a href="https://agentmods.dev/instructions/ctrixin/agent-2-agent/claude-md"><img src="https://agentmods.dev/badge/instructions/ctrixin/agent-2-agent/claude-md.svg" alt="Measured on agentmods" 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 | $0.00659 | $0.00659 |
| Opus 5 | $0.00329 | $0.00329 |
| Sonnet 5 | $0.00132 | $0.00132 |
| Haiku 4.5 | $0.00066 | $0.00066 |
Grade A, and why
agent-2-agent CLAUDE.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 4d 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.
What it actually says
a2a — AI Agent 快速上下文
本文件供 AI agent 首次进入项目时快速理解全局，人类开发者请看 README.md。
这是什么
a2a 是一个 Claude Code skill（不是独立应用），实现跨模型对抗式 code review。核心逻辑全在 SKILL.md，脚本只做辅助。
文件职责
| 文件 | 作用 | 什么时候改 |
|---|---|---|
SKILL.md |
skill 定义 + 完整执行流程，Claude Code 加载的入口 | 改流程、改 prompt、改 review 逻辑 |
scripts/preflight.sh |
环境预检 + token bridge 缓存，结果缓存 15min | 改检测逻辑、加新 CLI 支持 |
scripts/a2a-health.sh |
一键自检 + 自动修复 token bridge，人工排查用 | 改检测项、改修复逻辑 |
scripts/spinner.sh |
终端动画，纯 UI | 基本不用动 |
references/review-lenses.md |
三大审查视角定义（Challenger/Architect/Subtractor） | 调整审查标准 |
DEV.md |
开发同步流程 + 版本号规则 | 流程变了才改 |
关键设计决策
- 双向跨模型 dispatch：Claude→Codex 走
codex exec；Codex→Claude 走claude -p+ token bridge（或从 Claude Code 内走 Agent tool sub-agent）。 - Token Bridge：OAuth token 存在 macOS Keychain，Codex sandbox 无法直接访问。preflight 自动提取 token 缓存到
~/.cache/a2a/.claude-token（权限 600），Codex 通过ANTHROPIC_API_KEY环境变量传递。Step 1 每次自动检查补全，用户零操作。 - 从 Claude Code 内不直接调
claude -p：CLAUDECODE=1环境变量触发 nested session 保护。Claude 侧 review 走 Agent tool。 - Reviewer 互不可见：每个 reviewer 独立执行，看不到其他 reviewer 的输出，防止从众效应。
- Lean packet：reviewer 只收到自己视角需要的 diff 切片 + 精简 intent，不传完整 plan 或项目规范。
- Red-line 统一扫描：项目约束不传给 reviewer，由 Claude 在 Step 5 汇总时统一扫描一次。
修改注意事项
- 改完必须同步更新
CHANGELOG.md+SKILL.md里的 version - 同步到开源 repo 流程见
DEV.md - Cache 路径：
~/.cache/a2a/（preflight.json + .claude-token），调试时用preflight.sh --refresh或直接跑a2a-health.sh
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.
- 4d ago First seen · 34 lines · 659 tokens per session scan A 9373735cb493
agent-2-agent CLAUDE.md is an instructions file published in the GitHub repository CtriXin/agent-2-agent (9 stars, last pushed 5mo ago), licensed MIT. It adds 659 tokens to every session, about $0.0033 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
harness-kit CLAUDE.md
Claude Code instructions for harnessprotocol/harness-kit, covering harness-kit, repo structure, adding a new plugin, 1. create the plugin directory and 2. write the plugin manifest.
skills AGENTS.md
Instructions for citedy/skills, covering agent instructions — citedy skills, repository, directory structure, adding a new skill and skill types.
FableClaudeMDForOpus CLAUDE.md
Claude Code instructions for altafino/FableClaudeMDForOpus, covering iron rules, project and hard stops.
skill-vault CLAUDE.md
Instructions for Hainrixz/skill-vault, covering skill vault — vault master agent, project structure, key files, commands and /vault-add — add a skill to the vault.
AmbyKit CLAUDE.md
Instructions for ambystechcom/AmbyKit, covering claude code — notes for the ambykit repo and claude-specific.
skillsync-mcp CLAUDE.md
Instructions for adityasugandhi/skillsync-mcp, covering claude.md, project overview, development commands, architecture and key files.