Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/nidhish28guhan-netizen/hachiman-agentnpx agentmods add skills/nidhish28guhan-netizen/hachiman-agent/skillWrote 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/skills/nidhish28guhan-netizen/hachiman-agent/skill)<a href="https://agentmods.dev/skills/nidhish28guhan-netizen/hachiman-agent/skill"><img src="https://agentmods.dev/badge/skills/nidhish28guhan-netizen/hachiman-agent/skill/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.
<a href="https://agentmods.dev/skills/nidhish28guhan-netizen/hachiman-agent/skill"><img src="https://agentmods.dev/badge/skills/nidhish28guhan-netizen/hachiman-agent/skill.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00000 | $0.00946 |
| Opus 5 | $0.00000 | $0.00473 |
| Sonnet 5 | $0.00000 | $0.00189 |
| Haiku 4.5 | $0.00000 | $0.00095 |
Grade A, and why
skill 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 12d 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.
HACHIMAN — Offensive Security Skill (AI-builder instructions)
Layer 1 of the Master Security Skill Architecture: this is what an AI coding agent (Claude Code, Codex, Cursor, Hermes, DeepSeek Harness, Qoder, …) must know to operate Hachiman's offensive loop against an authorized target. Keep this file next to the repository.
What this skill does
Hachiman is a watchman that thinks like an attacker. On an authorized target it runs:
DISCOVER → MAP → HYPOTHESIZE → ATTACK → ADAPT → CHAIN → VALIDATE → EXPLAIN → FIX → RETEST
Signature loop: ATTACK → PROVE → UNDERSTAND → FIX → REATTACK → VERIFY.
It is not a scanner that emits warnings. It produces confirmed, reproduced findings, an
AI Repair Contract per finding, and an independent retest verdict (VERIFIED,
UNRESOLVED, REGRESSION) for every fix.
Non-negotiable rules of engagement (hard-coded)
- Authorization first. Every run needs an engagement file with
authorized_byset. No authorization → Hachiman refuses to run. This is enforced in code, not convention. - Scope is a wall. Tools outside
scope.allowed_toolsare never touched; violations abort the run and are audit-logged. - No persistence. No real exfiltration. No destructive testing. (doc-06 §46 defaults.) Exfil demonstrations are allowed only against the local canary sink when explicitly enabled.
- Budgets are enforced: requests, duration, concurrency. Exhausting a budget stops the run and reports partial results honestly.
Operating commands (run from the repository root)
# 1. Authorize + scope the target (write an engagement JSON — see examples/)
node bin/hachiman.js pentest examples/engagement.vuln-notes.json
# 2. Recon only (no attacks)
node bin/hachiman.js recon examples/engagement.vuln-notes.json
# 3. Inspect results
node bin/hachiman.js findings
node bin/hachiman.js explain <finding-id> # root-cause developer card
node bin/hachiman.js fix <finding-id> # AI Repair Contract (yaml + json)
node bin/hachiman.js chain --eng <engagement> # attack graph + chained impact
node bin/hachiman.js report pentest <engagement>
# 4. After YOUR fix: prove it
node bin/hachiman.js retest <finding-id> --fixed <fixture|url>
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.
- 12d ago First seen · 81 lines · 0 tokens per session scan A 76a665dcbd9c
skill is a skill published in the GitHub repository nidhish28guhan-netizen/hachiman-agent (1 stars, last pushed 19d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 946 tokens. 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 2026-08-31.
Other skills, from other repositories
detecting-ai-model-prompt-injection-attacks
Detects prompt injection attacks targeting LLM-based applications using a multi-layered defense combining regex pattern matching for known attack signatures, heuristic scoring for structural anomalies, and transformer-based classification with DeBERTa models. The detector analyzes user inputs before they reach the…
panguard
AI agent security platform — audit skills, scan for threats, and run 24/7 protection with 9,700+ detection rules.
codex-multi-agent-director
Mission: decompose non-trivial HELM implementation or audit work into bounded, reviewable task threads.
repo-auditor
Mission: produce a source-grounded reality audit before implementation work.
clawmoat
Real-time AI agent security scanner. Detects prompt injection, jailbreak attempts, credential/secret leaks, PII exposure, and dangerous tool calls. Activate when: (1) scanning inbound messages or tool outputs for prompt injection, (2) checking outbound content for credential leaks or PII, (3) auditing agent session…
subagent-driven-development
Use when executing implementation plans with independent tasks in the current session.