agentmods.dev

OpenHack AGENTS.md

A required process for using coding agents to perform penetration tests, which are controlled checks for security weaknesses in software.

In plain words
What is it for?
Use it to start or resume a security run, choose security experts, perform reconnaissance, optionally scan for code patterns and create a prioritized work list.
Why use it?
It prevents an agent from starting an unstructured security review and keeps expert selection, findings, decisions and generated files organized.

Instructions file for CodexOpenCode

735 3mo ago A scan 2,001 tokens original MIT
View the source file hadriansecurity/OpenHack
Install

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.

agentmods
npx agentmods add instructions/hadriansecurity/openhack/agents-md
Clone the repo
git clone --depth 1 https://github.com/hadriansecurity/OpenHack

Made for: Codex, OpenCode.

Per session 2,001 This file is loaded in full into every session.
When invoked 2,001 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.02001 $0.02001
Opus 5 $0.01001 $0.01001
Sonnet 5 $0.00400 $0.00400
Haiku 4.5 $0.00200 $0.00200

Measured yesterday against content hash 273a706fc0ba, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

OpenHack 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.

AGENTS.md · 150 lines

How it starts

The opening of the file, as written. The whole thing — 150 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Instructions

This repository is a file-based openhack workspace. When a human asks to start, initiate, run, resume, or continue a pentest/security review, use the tool structure before doing vulnerability analysis.

Required Pentest Flow

Do not begin with a broad LLM source sweep, direct expert pass, or ad hoc manual review of the target repository.

For a new target, use phase checkpoints:

  1. Run openhack init-run <target> <git-url> [--run-id <run-id>] [--branch <branch>].
  2. Summarize the created run and source commit, then ask the human what security experts to use before recon. Show every configured expert and the option all agents.
  3. Run openhack run-recon <target> <run-id> --all-agents only after approval, or run it with one or more --expert <expert-id> options if the human selected a subset.
  4. Summarize recon counts and artifacts; ask whether to proceed to scenario routing and backlog generation. If the human wants deeper source-pattern coverage, rerun recon with --semgrep before scenario routing. Recon writes routing-units.jsonl; treat those clustered units as the primary router input instead of asking the router to reason over the full raw inventory dump.
  5. After approval, run openhack create-scenarios <target> <run-id>, have the scenario-router answer the rendered prompt, and record that router output with openhack record-scenario-backlog <target> <run-id> router-result.json without inserting another human confirmation between prompt creation and backlog recording. The scenario-router must use the expert scope recorded before recon; do not create scenarios for unselected experts. The router must route from mandatory routing_unit_id + expert pairs when routing units are present, and use path/expert coverage requirements as a compatibility backstop.
  6. Summarize backlog size and coverage notes; ask for one approval to run the entire unfinished scenario backlog. Batch approval is not batch analysis: every scenario still needs its own rendered prompt, source review, evidence, and result.
  7. Review only recorded runs/<target>/<run-id>/scenarios/backlog/S*.md expert prompts and record results with openhack record-scenario-result <target> <run-id> .... This records finished scenario results and finding-candidates/; it does not create final findings/.
  8. Summarize finding-candidate count and ask once to run the entire unfinished finding triage backlog. Every candidate must use its own rendered runs/<target>/<run-id>/finding-triage/prompts/S###-F###.md prompt and its own finding-triage subagent. Render prompts with openhack render-finding-triage-prompt <target> <run-id> <candidate-id>.
  9. Record each finding-triage agent answer with openhack record-finding-triage <target> <run-id> <candidate-id> triage-result.json. Only accepted and downgraded triage decisions may materialize final findings/.
  10. Validate with openhack validate-run <target> <run-id>.

Read the full file on GitHub · 150 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. yesterday First seen · 150 lines · 2,001 tokens per session scan A 273a706fc0ba

Subscribe to this mod's changes

OpenHack AGENTS.md is an instructions file published in the GitHub repository hadriansecurity/OpenHack (735 stars, last pushed 3mo ago), licensed MIT. It adds 2,001 tokens to every session, about $0.0100 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.

Data and AISecurityTestingcybersecurityhacking
Related

Other instructions, from other repositories

networking-with-go AGENTS.md

AGENTS.md instructions for sazardev/networking-with-go, covering agents.md, repo identity, commands, running exercises and critical rules.

sazardev/networking-with-go · 826 tokens

networking-with-go CLAUDE.md

Claude Code instructions for sazardev/networking-with-go, covering claude.md, what this repo is, repository structure, running exercises and writing/editing documentation chapters.

sazardev/networking-with-go · 746 tokens

redamon AGENTS.md

Instructions for samugit83/redamon, covering redamon - agent ruleset (repository root), auto-invoke skills, critical rules - non-negotiable, code comments and tech stack.

samugit83/redamon · 2,276 tokens

stride-gpt AGENTS.md

Instructions for mrwadams/stride-gpt, covering agents.md, what this project is, repository layout, how the agentic analysis works and progressive disclosure pattern.

mrwadams/stride-gpt · 4,122 tokens

DorkAgent CLAUDE.md

Instructions for yee-yore/DorkAgent, covering claude.md, project overview, commands, running the tool and interactive mode (menu-driven interface).

yee-yore/DorkAgent · 1,888 tokens

Huntable-CTI-Studio AGENTS.md

Instructions for dfirtnt/Huntable-CTI-Studio, covering agents.md, orientation, local context, prompt-injection alerting and change-type quick reference.

dfirtnt/Huntable-CTI-Studio · 3,724 tokens