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/kungfu-systems/kungfu/agents-mdgit clone --depth 1 https://github.com/kungfu-systems/kungfuWhat 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 | $0.01532 | $0.01532 |
| Opus 5 | $0.00766 | $0.00766 |
| Sonnet 5 | $0.00306 | $0.00306 |
| Haiku 4.5 | $0.00153 | $0.00153 |
Grade A, and why
kungfu 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 yesterday.
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 — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file orients coding agents (and people) working with this repository. It is a router, not a duplicate: it points to the authoritative documents rather than restating them.
Start from the person's objective
Restate the requested outcome in one sentence, then choose the single matching route. Do not inventory the repository or preload multiple routes "just in case."
| Objective | First route |
|---|---|
| Use or operate an installed Kungfu | Read the curated docs/README.md; run kungfu agent brief, then kungfu agent choose-mode --json. |
| Explain or evaluate Kungfu as a whole | Follow the Agent Architecture Analysis Protocol, then use the Evolution Map to enter current authority. |
| Change one bounded component, behavior, or document | Read CONTRIBUTING.md, then compile the verified task context below. |
| Promote or publish a release | Load the version and release design and publication closure before acting. |
Use the exhaustive docs/MAP.md only to ground a specific claim
or locate authority not resolved by the selected route. Expand context when a
verified route reports an omission or the task actually crosses a boundary—not
because the repository contains more concepts.
Rules that always apply
- Treat checked-in contracts, source, and current verification output as authority. README text, chat history, generated projections, and Agent output are routes or evidence, not authority by themselves.
- Keep the person's mental model task-sized. Do not export implementation vocabulary, subsystem count, or repository structure unless it changes the answer or the person asks for it.
- Use
./shifufor every development, build, test, package, and documentation task. Do not invoke pnpm, node, conan, or cmake directly. - Prefer the smallest change that preserves existing behavior and authority; read the affected source and tests before editing.
- Never include secrets, credentials, tokens, or private logs in code, commits,
issues, or pull requests. Follow
SECURITY.mdfor private vulnerability reporting. - Do not create official integrations that scrape private provider sessions,
bypass billing or quotas, or misrepresent usage attribution. See
PROVIDER_COMPLIANCE.md.
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.
- yesterday First seen · 121 lines · 1,532 tokens per session scan A 0854af1b241c
kungfu AGENTS.md is an instructions file published in the GitHub repository kungfu-systems/kungfu (4,505 stars, last pushed yesterday), licensed Apache-2.0. It adds 1,532 tokens to every session, about $0.0077 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
build_a_product_agent CLAUDE.md
Instructions for ADW-19/build_a_product_agent, covering claude.md, 项目本质（readme 概览之外）, 目录与编号规范, 文章写作规范（新增/修改文章时必须遵循） and 常用操作.
pragma AGENTS.md
AGENTS.md instructions for pqpo/pragma, covering agents.md, 项目定位, 技术基线, 目录结构 and package 命名.
beevibe CLAUDE.md
Instructions for beevibe-ai/beevibe, covering beevibe — claude code guide, monorepo structure, running locally, critical: @beevibe/core compiles to dist/ and restarting cleanly.
pi-agent-dashboard AGENTS.md
Instructions for BlackBeltTechnology/pi-agent-dashboard, covering pi dashboard, docs-first gate — kb before grep (per-turn doctrine), code instructions (per-turn doctrine), investigation protocol — index first and documentation update protocol.
agents-universe CLAUDE.md
Instructions for agents-universe/agents-universe, covering agents universe — enterprise ai agent framework, project overview, monorepo layout, development commands and api (from packages/api/).
BuildersGate CLAUDE.md
Instructions for Thepizzapie/BuildersGate, covering setting up builders gate for a user, ask two things first, install, point it at a game and keys.