PentestGPT is an AI-driven framework for penetration testing and capture-the-flag challenges. It guides staged workflows such as reconnaissance, asset discovery, vulnerability identification, exploitation, and reporting, using large language models to operate tools and reason about findings. The catalogue includes skills and instructions for using it.
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/greydgl/pentestgpt/claude-mdgit clone --depth 1 https://github.com/GreyDGL/PentestGPTWrote 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/greydgl/pentestgpt/claude-md)<a href="https://agentmods.dev/instructions/greydgl/pentestgpt/claude-md"><img src="https://agentmods.dev/badge/instructions/greydgl/pentestgpt/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 | $0.00780 | $0.00780 |
| Opus 5 | $0.00390 | $0.00390 |
| Sonnet 5 | $0.00156 | $0.00156 |
| Haiku 4.5 | $0.00078 | $0.00078 |
Grade A, and why
PentestGPT 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 — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude guide
Internal instructions for Claude Code in this repository. Read
docs/architecture.md before architecture work. The root README.md is the
public project page and must not be edited unless the user explicitly requests that separate review.
Active project map
pentestgpt_agent/— maintained autonomous Supervisor/Executor framework; nested uv project.pentestgpt_legacy/— maintained human-driven USEN-2024-style client.../UnifedAgentWrapper/— canonicalunified-agentpackage used by the framework.../xbow-benchmark/— reference-only benchmark harness and historical results; it is not a supportedpentestgpt-agentCLI, CI, or runtime integration.unified_agent/— obsolete compatibility copy retained by root packaging/Docker only. Do not add features here or confuse it with the framework dependency.
The old fixed-stage pentestgpt/ package and the previous ledger-centered
Instructor/Executor/Judge implementation are gone. Historical reports may still name them.
Runtime contract
- The Supervisor and Executor both use
SandboxPolicy.FULL_ACCESS; deployment isolation is the security boundary. - Each episode is fresh. SQLite and exact trace receipts—not provider conversation history—are memory.
- Deterministic code owns scope, plan validation, leases, evidence, retries, completion bases, and revisions.
- Provider action and file activity is logged but is not an audit failure.
- Keep the core at two LLM roles. Do not add an always-on judge, RAG layer, or scheduler without trace evidence that the current deterministic seam cannot solve the problem.
Commands
Framework work runs inside the nested project:
cd pentestgpt_agent
uv sync --extra claude # or codex / all
uv run python -m pytest -q
uv run ruff check src tests
uv run ruff format --check src tests
uv run mypy src
uv lock --check
uv build
From the repository root:
make run TARGET=http://127.0.0.1:8000 BACKEND=claude
make check
make ci
make docker-build
make docker-login
make docker-auth-status
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 · 81 lines · 780 tokens per session scan A 31c3a605293a
PentestGPT CLAUDE.md is an instructions file published in the GitHub repository GreyDGL/PentestGPT (15,220 stars, last pushed 1mo ago), licensed MIT. It adds 780 tokens to every session, about $0.0039 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
LightRAG AGENTS.md
AGENTS.md instructions for HKUDS/LightRAG, covering repository guidelines, project overview, project structure, module layout (lightrag/) and core architecture.
agentscope copilot-instructions.md
Copilot instructions for agentscope-ai/agentscope, covering agentscope code review guide, 1. code quality, 2. [must] code security, 3. [must] testing & dependencies and 4. code standards.
redis-vl-python CLAUDE.md
Claude Code instructions for redis/redis-vl-python, covering claude.md - redisvl project context, frequently used commands, development workflow, redis setup and documentation.
awesome-code-agents CLAUDE.md
Claude Code instructions for EuniAI/awesome-code-agents, covering awesome code agents — project context, working preferences (owner: zhaoyang-chu), what this is, the two sources of truth and data layout.
langfuse AGENTS.md
AGENTS.md instructions for langfuse/langfuse, a project described as: 🪢 Open source AI engineering platform: LLM evals, observability, metrics, prompt management, playground, datasets. Integrates with OpenTelemetry, LangChain, OpenAI SDK, LiteLLM, and more. 🍊YC W23.
langfuse CLAUDE.md
Claude Code instructions for langfuse/langfuse, a project described as: 🪢 Open source AI engineering platform: LLM evals, observability, metrics, prompt management, playground, datasets. Integrates with OpenTelemetry, LangChain, OpenAI SDK, LiteLLM, and more. 🍊YC W23.