token-ninja: Instructions file for Claude Code

CLAUDE.md

token-ninja CLAUDE.md is an instructions file for Claude Code from oanhduong/token-ninja. It costs 1,736 tokens per session, scanned C, original, MIT.

Repository instructions for Token Ninja, a command-line tool and MCP server that recognizes safe, deterministic requests and runs them locally, while passing other requests to an AI coding assistant.

In plain words
What is it for?
Working on command routing, safety validation, rule classification, local execution, AI fallback, setup and diagnostics, and the related test suite.
Why use it?
They explain the project layout, routing behavior, safety checks, commands, and test expectations so changes do not bypass its safety or fallback rules.

Instructions file for Claude Code

Written for Claude Code: UserPromptSubmit hook event. Also seen: mentions CLAUDE.md; mentions Claude Code; mentions Codex.

This is oanhduong/token-ninja's own configuration. It tells Claude Code how to work on token-ninja itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything token-ninja configures →

Reuse

Borrowing it

Nothing to install: this file belongs to oanhduong/token-ninja. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/oanhduong/token-ninja/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/oanhduong/token-ninja

Made for: Claude Code.

Wrote 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.

agentmods badge for token-ninja CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/oanhduong/token-ninja/claude-md/github.svg)](https://agentmods.dev/instructions/oanhduong/token-ninja/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/oanhduong/token-ninja/claude-md"><img src="https://agentmods.dev/badge/instructions/oanhduong/token-ninja/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for token-ninja CLAUDE.md

Your own site · 80×15
<a href="https://agentmods.dev/instructions/oanhduong/token-ninja/claude-md"><img src="https://agentmods.dev/badge/instructions/oanhduong/token-ninja/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 1,736 This file is loaded in full into every session.
When invoked 1,736 The same file — it is already loaded in full.
Security scan C 2 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.01736 $0.01736
Opus 5 $0.00868 $0.00868
Sonnet 5 $0.00347 $0.00347
Haiku 4.5 $0.00174 $0.00174

Measured 9d ago against content hash 32def498a562, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade C, and why

token-ninja CLAUDE.md scanned grade C with 2 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 9d 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.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

denylist.ts # DENY_PATTERNS regex list (rm -rf, sudo, curl|sh, etc.)

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

denylist.ts # DENY_PATTERNS regex list (rm -rf, sudo, curl|sh, etc.)
CLAUDE.md · 139 lines

How it starts

The opening of the file, as written. The whole thing — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Notes for Claude Code (and other AI coding assistants) working inside this repository.

What this project is

token-ninja is a CLI + MCP server that intercepts commands meant for an AI coding assistant and runs the deterministic ones locally — zero LLM calls. Anything it doesn't confidently recognize falls back to the user's AI tool (Claude, Codex, Cursor, Aider, Gemini, Continue).

  • 765 built-in rules across 46 tool domains
  • Classifier hot path: ~19 µs/call; safety validator: ~10 µs/call (warm JIT)
  • 312 tests across 21 test files; 90.67% lines / 84.61% branches / 95.45% functions (v8 coverage over src/router/**, src/safety/**, src/rules/**; thresholds enforced at 88 / 82 / 88 / 95)

Source layout

src/
  cli.ts                     # commander entry; subcommands: mcp, route, setup, uninstall, init, stats, doctor, shim, rules
  doctor/
    index.ts                 # `ninja doctor` health check (config, rules, shim, MCP, hook, stats)
  router/
    index.ts                 # runRouter(): safety → classify → safety (again) → exec-or-fallback
    route-once.ts            # one-shot classify+exec used by the Claude Code UserPromptSubmit hook (no AI fallback)
    classifier.ts            # exact → prefix → regex → nl matching; template expansion
    executor.ts              # execa wrapper (shell:true, captures + streams)
    fallback.ts              # hand off to user's AI tool
  rules/
    loader.ts                # loads builtin/*.yaml + ~/.config/token-ninja/rules/*.yaml
    types.ts                 # Rule / MatchSpec / ActionSpec
    builtin/*.yaml           # 765 rules, grouped by domain (46 files)
  safety/
    denylist.ts              # DENY_PATTERNS regex list (rm -rf, sudo, curl|sh, etc.)
    validator.ts             # pipeline split + NFKC + homoglyph strip, then match
  adapters/                  # per-AI-tool metadata + shim generator (claude-code, codex, cursor, aider, gemini, continue, generic)
  setup/
    index.ts                 # orchestrates `ninja setup`: shell shim + MCP register + hook install
    shell-install.ts         # writes managed block into ~/.bashrc / ~/.zshrc
    mcp-install.ts           # registers ninja mcp with Claude Code, Cursor, Claude Desktop, Gemini CLI
    hook-install.ts          # installs Claude Code UserPromptSubmit hook
  mcp/server.ts              # stdio MCP server exposing tool: maybe_execute_locally
  telemetry/stats.ts         # hit/miss counters, tokens-saved estimate
  config/user-config.ts      # ~/.config/token-ninja/config.yaml
  utils/
    shell-parse.ts           # splitPipelineSegments, tokenize, normalizeNl
    repo-detect.ts           # detect markers (package.json, Cargo.toml, …) and PMs
    logger.ts                # ANSI-colored stderr
hooks/
  claude-code-user-prompt.cjs # shipped hook: reads a prompt, calls `ninja route`, short-circuits the model on hit
tests/                       # 20 test files, 268 tests, vitest + v8 coverage
  fixtures/real-commands.txt # ≥85% of these must classify (rules-coverage.test.ts)

Read the full file on GitHub · 139 lines

Changes

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.

  1. 9d ago First seen · 139 lines · 1,736 tokens per session scan C 32def498a562

Subscribe to this mod's changes

token-ninja CLAUDE.md is an instructions file published in the GitHub repository oanhduong/token-ninja (38 stars, last pushed 4mo ago), licensed MIT. It adds 1,736 tokens to every session, about $0.0087 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other instructions, from other repositories

next.js AGENTS.md

AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.

vercel/next.js · 7,296 tokens

codex AGENTS.md

AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.

openai/codex · 5,153 tokens

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

vscode oss-third-party-notices.instructions.md

Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).

microsoft/vscode · 5,001 tokens

langchain AGENTS.md

AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.

langchain-ai/langchain · 4,469 tokens

deepseek-harness AGENTS.md

AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.

deepseek-ai/deepseek-harness · 3,735 tokens