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/lookfree/cc-harness/claude-mdgit clone --depth 1 https://github.com/lookfree/cc-harnessWrote 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/lookfree/cc-harness/claude-md)<a href="https://agentmods.dev/instructions/lookfree/cc-harness/claude-md"><img src="https://agentmods.dev/badge/instructions/lookfree/cc-harness/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 | $0.03406 | $0.03406 |
| Opus 5 | $0.01703 | $0.01703 |
| Sonnet 5 | $0.00681 | $0.00681 |
| Haiku 4.5 | $0.00341 | $0.00341 |
Grade B, and why
cc-harness CLAUDE.md scanned grade B with 1 finding 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 5d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- **plugin 是否启用** → `~/.claude/settings.json` 的 `enabledPlugins`（plugin 级，不是版本级）。 How it starts
The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cc-harness
围绕 Claude Code 的桌面工具。Electron + React + TypeScript，双模式（桌面 Electron / Web Express）。
定位正在演进：从一年前的"Claude Code 配置浏览器"重做成 "Harness 工作台"——配置 / 调试 / 观测 / 编排 / 教学 五件事。详细方向、路径、按版本对齐的功能缺口、以及每个功能的可执行 spec，全部在 docs/：
docs/cc-harness演进路径.md—— 产品方向、五支柱、Phase 0/1/2/3 划分（先读这个）。docs/harness-ide-spec/README.md—— 22 个实现 spec 的索引（spec001–022）+ 状态 + 依赖关系。docs/harness-ide-spec/功能版本对照表.md—— 功能 ID（ORCH/OBS/HOOK/PERM/MODEL/SKILL/MISC）↔ Claude Code 版本 ↔ 项目覆盖状态。
动代码前先看对应 spec（spec 引了真实 file:line、给了类型 diff 和验收标准）。改了方向/功能后回头更新这三份文档。
工作方式
- 想清楚再写：先说清假设；有多种理解就摆出来让我选，别默默挑一个；有更简单的做法就直说、该反对就反对；不清楚就停下来问，别猜。
- 最简实现：解决问题的最少代码，不加没要求的功能/抽象/配置/防御。单次使用就别造抽象。写完若 200 行能压到 50，就重写。和项目"不做什么"哲学一致（Thin Harness）。
- 外科手术式改动：只动该动的，每一行改动都能追溯到需求。别顺手"改进"相邻代码/注释/格式，别重构没坏的东西，跟现有风格走。只清理自己改动产生的孤儿（import/变量）；遇到无关的既有死代码——提一句，别删（除非明确要求）。
- 目标驱动：把任务变成可验证的成功标准再做。多步任务先列简短计划，每步带验证点（改 X → 验证：Y）。spec 的"验收标准"就是现成的成功标准，照着 loop 到通过。
跑起来
npm install
npm run electron:dev # 桌面端（主模式）
npm run web:dev # Web 模式（Express :3001 + Vite :5173）
npm run electron:build # 出包
⚠️ 已知 build 时序 bug（spec001 修）：electron:dev 里 copy:preload 可能在 vite 首次 build 出 dist-electron/ 之前就跑，cp 失败 → preload 没拷过去 → window.electronAPI 全断、IPC 不通。首次启动若 Electron 窗口空白/报 Preload error，先 mkdir -p dist-electron && cp electron/preload.cjs dist-electron/preload.cjs 再重启。彻底修见 spec001。
架构
electron/services/file-manager*.ts # 核心：扫描/读写 ~/.claude 和项目 .claude 配置。
# 按域拆成继承链（base→plugins→skills→agents→hooks→mcp→commands→claudemd→FileManager）；
# FileManager 仍是单例门面，对外 API 不变。新增方法放对应 domain 文件，跨域共享的低层 helper 放 base（protected）。
electron/services/provider-manager.ts
electron/ipc/*.ts # 每个域一个 registerXxxHandlers(ipcMain, fileManager)
electron/preload.cjs # contextBridge（必须 .cjs）
server/index.ts # Web 模式 REST，镜像 IPC
src/pages/*.tsx # Dashboard/Skills/Agents/Hooks/MCP/Commands/ClaudeMd/Graph/Models/Settings
src/lib/api.ts # 统一 API，自动探测 Electron(IPC) / Web(HTTP)
shared/types/*.ts # 主进程/渲染进程共享类型
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.
- 5d ago First seen · 82 lines · 3,406 tokens per session scan B aa61d090e129
cc-harness CLAUDE.md is an instructions file published in the GitHub repository lookfree/cc-harness (47 stars, last pushed 1mo ago), licensed MIT. It adds 3,406 tokens to every session, about $0.0170 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
kungfu copilot-instructions.md
Copilot instructions for kungfu-systems/kungfu: Read and follow AGENTS.md before proposing or running changes. It is the shared agent router for this repository, including the canonical Shifu development and build entrypoint.
agents CLAUDE.md
Instructions for eloylp/agents, covering claude.md, project overview, directory structure, config model and build & run.
kubeledger CLAUDE.md
Claude Code instructions for realopslabs/kubeledger, covering claude.md, project overview, development commands, dependency management and install dependencies.
ideate AGENTS.md
Instructions for paultyng/ideate, covering agents.md, build features as mcp tools first, ui second, when to add ui and pattern.
skim CLAUDE.md
Instructions for bb1nfosec/skim, covering project, response rules and forbidden — never read.
icron CLAUDE.md
Instructions for zebbern/icron, covering icron, architecture, key directories, built-in tools and commands.