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/nanoagentteam/research-claw/claude-mdgit clone --depth 1 https://github.com/nanoAgentTeam/research-clawWrote 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/nanoagentteam/research-claw/claude-md)<a href="https://agentmods.dev/instructions/nanoagentteam/research-claw/claude-md"><img src="https://agentmods.dev/badge/instructions/nanoagentteam/research-claw/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 5.1 | $0.00833 | $0.00833 |
| Opus 5 | $0.00417 | $0.00417 |
| Sonnet 5 | $0.00167 | $0.00167 |
| Haiku 4.5 | $0.00083 | $0.00083 |
Grade A, and why
research-claw 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 6d 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
CLAUDE.md
这是一个为 Claude Code (claude.ai/code) 在此存储库中工作时提供指导的文件。
重要提示:请始终使用简体中文回答用户的问题和进行交流。
代码要兼容 python 3.11
开发指令
核心执行
- 交互式 Agent:
python cli/main.py agent - 单次消息 Agent:
python cli/main.py agent -m "你的指令" -p ProjectID - 网关模式 (飞书/Telegram):
python cli/main.py gateway --port 18790 - 启动子 Agent:
python cli/main.py subagent --role RoleName --project-id PID --session SID - 初始配置:
python cli/main.py onboard(创建配置文件与工作区)
环境配置
python 3.11+
- 安装依赖:
pip install -r requirements.txt - 认证: Overleaf 同步需要
.olauth文件 (通过ols login生成)。
代码架构
概览
Research Claw 是一个自托管的 AI 科研助手,管理论文项目、检索文献、追踪学术进展与截稿日期,并通过多渠道随时响应。它基于分层 AI Agent 架构,管理一个 workspace/ 目录,分为 Default (全局管理) 与 Project (特定论文) 空间。
核心组件 (core/)
Project(core/project.py): 核心抽象。管理core目录 (LaTeX 文件)、Git 版本控制、Overleaf 同步以及 LaTeX 编译。Session(core/session.py): 在项目内管理独立的对话历史、元数据以及子 Agent 的工作区。FS(core/fs/): 提供路径安全的文件操作以及为子 Agent 提供 Overlay 虚拟文件系统。Tools(core/tools/): 模块化的工具系统 (搜索、编译、Git 等),通过config/tools.json配置。
Agent 系统 (agent/)
AgentLoop(agent/loop.py):Assistant角色的主执行循环。处理工具调用与指令路由。WideAgentLoop(agent/wide_loop.py): 编排复杂的多 Agent 任务 (Swarm/DAG 模式)。GitAgent(agent/git_agent.py): 专门用于版本控制操作的 Agent。
分层与隔离
- Assistant (主 Agent): 对项目的
core/目录拥有完整的读写权限。 - Worker (子 Agent): 在 Session 特定的
subagents/覆盖层中工作。变更仅在Assistant核准后或通过MergeAgent合并至core/。
工作区结构
workspace/{项目名}/{项目名}/: "core" 目录 (实际的 LaTeX 文件 + Git)。workspace/{项目名}/{MMDD_NN}/: Session 目录,用于历史记录与隔离。workspace/{项目名}/project.yaml: 项目特定设定 (Overleaf ID, Git 配置, LaTeX 引擎)。
开发指南
- 路径安全: 务必使用
project.resolve(path)或session.fs以防止路径穿越 (Path Traversal)。 - 日志: 使用
loguru进行结构化日志记录。 - 配置: 全局设置位于
settings.json;功能开关位于config/*.json。
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.
- 6d ago First seen · 61 lines · 833 tokens per session scan A 3dfd265a5a90
research-claw CLAUDE.md is an instructions file published in the GitHub repository nanoAgentTeam/research-claw (292 stars, last pushed 2mo ago), licensed MIT. It adds 833 tokens to every session, about $0.0042 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 2026-08-30.
Other instructions, from other repositories
harness-sdk AGENTS.md
AGENTS.md instructions for strands-agents/harness-sdk, covering agent development guide - strands agents monorepo, context, monorepo layout, where the "why" lives: team/ and writing code.
matryca-plumber CLAUDE.md
Claude Code instructions for MarcoPorcellato/matryca-plumber, covering code audit — code intelligence, always do, never do, resources and cli.
tokenjam AGENTS.md
AGENTS.md instructions for Metabuilder-Labs/tokenjam, covering agents.md — tokenjam and quick rules for agents that don't read claude.md.
pydantic-ai-todo CLAUDE.md
Instructions for vstorm-co/pydantic-ai-todo, covering claude.md, what this project is, commands, structure and core pattern.
hermes-desktop CLAUDE.md
Claude Code instructions for fathah/hermes-desktop, covering before starting work, post-task checklist (required — do not skip), what is lat.md?, commands and syntax primer.
spine CLAUDE.md
Claude Code instructions for synaptixs/spine, covering spine — working on this repo, layout — where things live, invariants — break these and things get subtly wrong, gotchas that have bitten and before pushing.