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/coff0xc/ctf-mcp/claude-mdgit clone --depth 1 https://github.com/Coff0xc/CTF-MCPWrote 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/coff0xc/ctf-mcp/claude-md)<a href="https://agentmods.dev/instructions/coff0xc/ctf-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/coff0xc/ctf-mcp/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 | $0.01460 | $0.01460 |
| Opus 5 | $0.00730 | $0.00730 |
| Sonnet 5 | $0.00292 | $0.00292 |
| Haiku 4.5 | $0.00146 | $0.00146 |
Grade A, and why
CTF-MCP 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.
How it starts
The opening of the file, as written. The whole thing — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
CTF-MCP is an MCP (Model Context Protocol) server that provides 140+ tools for CTF (Capture The Flag) challenges across Crypto, Web, Pwn, Reverse, Forensics, and Misc categories. It also includes an autonomous solving orchestrator and LLM agent integration.
Commands
# Install (editable mode)
pip install -e .
# Install with dev tools
pip install -e ".[dev]"
# Install with crypto/web deps
pip install -e ".[standard]"
# Install everything (Linux-only deps like pwntools excluded on Windows)
pip install -e ".[full]"
# Install LLM agent support
pip install -e ".[llm]"
# Run MCP server
ctf-mcp
# Run all tests
pytest
# Run a single test file
pytest tests/test_crypto.py
# Run a specific test
pytest tests/test_crypto.py::TestCryptoBase64::test_base64_encode -v
# Run with coverage
pytest --cov=ctf_mcp --cov-report=term-missing
# Lint
ruff check ctf_mcp/
black --check ctf_mcp/
# Format
black ctf_mcp/
ruff check --fix ctf_mcp/
# Security scan
bandit -r ctf_mcp/ -c pyproject.toml
# Pre-commit hooks
pre-commit install
pre-commit run --all-files
Architecture
Layer Overview
server.py (MCP entry point — dynamic tool registration via introspection)
├── tools/ 6 tool classes exposed as MCP tools (prefix_method naming)
├── core/ Autonomous solving pipeline
├── engines/ Category-specific solving engines
├── adapters/ External binary tool wrappers
├── llm/ LLM-powered agent (ReAct pattern)
├── mcp/ MCP protocol extensions (streaming, tasks, registry)
├── network/ HTTP client, remote connections, exploit runner
└── utils/ Security decorators, input validation, helpers
Tool Registration System
server.py dynamically discovers and registers all tools at startup:
- Each
tools/*.pymodule has a class (e.g.,CryptoTools) with aget_tools()method returning{method_name: description} register_tools()iterates modules, introspects method signatures viainspect, and generates JSON Schema for MCP- Tools are registered as
{prefix}_{method_name}(e.g.,crypto_base64_encode,web_sql_payloads) - The
TOOL_REGISTRYdict maps tool names to(module_instance, method_name)tuples for dispatch
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 · 131 lines · 1,460 tokens per session scan A 7660ac66f23d
CTF-MCP CLAUDE.md is an instructions file published in the GitHub repository Coff0xc/CTF-MCP (24 stars, last pushed 5mo ago), licensed MIT. It adds 1,460 tokens to every session, about $0.0073 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
SecEng-Exercises CLAUDE.md
Claude Code instructions for fosres/SecEng-Exercises, covering claude.md, repository overview, repository structure, exercise categories and integer overflow exercises.
red-run CLAUDE.md
Claude Code instructions for blacklanternsecurity/red-run, covering red-run, engagement workflow, token budget, architecture and orchestrator variants.
fucking-awesome-web-security CLAUDE.md
Claude Code instructions for Correia-jpv/fucking-awesome-web-security, covering claude.md, what this repo is, files of note, editing rules and entry schema.
toolkit-mcp-server CLAUDE.md
Instructions for cyanheads/toolkit-mcp-server, covering developer protocol, what's next?, core rules, patterns and tool.
HOM-AIMOS AGENTS.md
AGENTS.md instructions for wallidsaydi-creator/HOM-AIMOS: This file is a public-tool compatibility entry point. It is not a second guidance authority.
claude-pentest-skills CLAUDE.md
Claude Code instructions for frendysanusi/claude-pentest-skills, covering critical safety rules, slash commands, /scope, /recon and /hunt.