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/episodeyu/harnesssmith/agents-mdgit clone --depth 1 https://github.com/EpisodeYu/HarnessSmithWhat 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 | $0.01981 | $0.01981 |
| Opus 5 | $0.00991 | $0.00991 |
| Sonnet 5 | $0.00396 | $0.00396 |
| Haiku 4.5 | $0.00198 | $0.00198 |
Grade A, and why
HarnessSmith 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 3d 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — HarnessSmith 快速上手
面向在本仓库里干活的 Agent 的导航地图：30 秒认清这是什么、代码在哪、怎么跑、怎么算"做完"。 这里只讲"怎么上手"；硬约束与红线看
CLAUDE.md，定位 / 范围 / 决策总表看docs/02-development/00-overview.md。两者冲突时以那两份为准。
1. 这是什么（一句话）
HarnessSmith 是 agent harness 的代码生成器（create-next-app for agent harnesses）：吃一份 HarnessSpec（YAML / preset / 向导采集），渲染出一个完整、独立、无 agent 框架锁定的 Python 仓库。生成的产物与本生成器零运行期关系——生成即脱离。
2. 两层心智（最重要，先记住）
任何改动，先问自己改的是哪一层：
| 层 | 位置 | 你在改什么 |
|---|---|---|
| 生成器本体 | harnessmith/*.py（spec / generator / cli / scaffold / wizard …） |
spec schema、渲染引擎、CLI、向导、catalog、preset |
| 产物模板 | harnessmith/templates/**/*.j2 |
渲染后才是用户拿到的代码。改这里 = 改所有未来产物 |
推论：测试必须覆盖"生成产物"本身，不能只测生成器。黄金路径 = 生成 → uv sync → pytest → mock LLM 跑通一次 function-calling（见 §5）。
3. 仓库地图
harnessmith/ # ← 生成器本体（这个包发布到 PyPI）
spec.py # HarnessSpec（Pydantic v2 + YAML，extra=forbid）
generator.py # 渲染模板 → 写仓库 + git init + uv lock + 冒烟自检
cli.py # Typer 入口：new / wizard / doctor
scaffold.py # 生成器与 CLI 向导共享的烤默认 / slug 派生（纯 stdlib）
cli_wizard.py # 终端交互向导（questionary）
node_bootstrap.py # Node 系 MCP server 的离线预热 / node 直跑
debuglog.py # 生成器侧 debug 日志
catalog/mcp_servers.yaml # 精选静态 MCP catalog（向导/CLI 预填数据源）
presets/coding-assistant/ # 内置 preset（spec.yaml + mcp_prefill.yaml）
wizard/ # Web 向导（FastAPI 单页，[wizard] extra，绝不进产物）
templates/ # ← 产物模板（.j2）。渲染出的才是用户的代码
src/__project_slug__/harness/ # 产物核心：loop/llm/llm_anthropic/tools/context/
# session/interaction/hooks/usage/trace/prompts/
# paradigms/ + mcp/skills/memory/subagents（opt-in）
src/__project_slug__/interfaces/ # cli.py（+ web.py / web_index.html opt-in）
tests/ # 产物自带测试（mock LLM）
docs/02-development/ # 设计与切片文档（00-overview = 唯一口径）
tests/ # ← 生成器自身的测试（test_spec / test_generator / test_golden …）
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.
- 3d ago First seen · 101 lines · 1,981 tokens per session scan A f1d332ae788f
HarnessSmith AGENTS.md is an instructions file published in the GitHub repository EpisodeYu/HarnessSmith (22 stars, last pushed 1mo ago), licensed MIT. It adds 1,981 tokens to every session, about $0.0099 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
apm python.instructions.md
Python development guidelines.
pysm AGENTS.md
Instructions for galsci/pysm, covering repository guidelines, project overview, project structure & key files, quick start commands and setup environment.
django-starter-template AGENTS.md
Instructions for wilfredinni/django-starter-template, covering agents.md, commands, lint & typecheck, architecture and testing.
foundry-local python-cffi-pointer-types.instructions.md
Use when working on the Python SDK's native bindings (cffi), especially when defining new Item subclasses, working with flItemQueue, or extending Request/Session methods that pass native handles.
llm-ide-rules pytest-tests.instructions.md
Instructions for iloveitaly/llm-ide-rules, covering pytest tests, example test and file structure.
easy-cheese python.instructions.md
Instructions for paulnsorensen/easy-cheese, covering hard rules, what to flag and what not to flag.