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/ShulkwiSEC/bb-hugenpx agentmods add skills/shulkwisec/bb-huge/ai-pair-hunting-with-claudeWrote 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/shulkwisec/bb-huge/ai-pair-hunting-with-claude)<a href="https://agentmods.dev/skills/shulkwisec/bb-huge/ai-pair-hunting-with-claude"><img src="https://agentmods.dev/badge/skills/shulkwisec/bb-huge/ai-pair-hunting-with-claude/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/shulkwisec/bb-huge/ai-pair-hunting-with-claude"><img src="https://agentmods.dev/badge/skills/shulkwisec/bb-huge/ai-pair-hunting-with-claude.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 | $0.00058 | $0.03190 |
| Opus 5 | $0.00029 | $0.01595 |
| Sonnet 5 | $0.00012 | $0.00638 |
| Haiku 4.5 | $0.00006 | $0.00319 |
Grade A, and why
ai-pair-hunting-with-claude 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- ai-pair-hunting-with-claude — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 339 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Pair Hunting with Claude
When to Use
- When setting up Claude Code CLI to autonomously test a bug bounty target.
- When running overnight / multi-hour autonomous hacking sessions.
- When managing scope and context across multiple bug bounty programs.
- When Claude is getting stuck in compaction loops or losing context mid-session.
- When you need Claude to stay strictly within program scope.
Prerequisites
- Claude Code CLI installed and authenticated
- A defined bug bounty target with program policy
- Separate workspace directory per target program
- Understanding of Claude's context window limitations
Core Concept: Claude as a Pair Hunter
"AI is not your replacement. It is your Pair Hunter — it brings determinism (accuracy) and speed." — Critical Thinking Podcast, Ep. 166
Claude excels at:
- Deterministic tasks: Testing every parameter in a 200-param API
- Speed: Fuzzing/enumerating faster than manual testing
- Pattern recognition: Spotting anomalies in large response sets
- Documentation: Auto-generating reports from findings
Claude struggles with:
- Creative intuition: "This feels wrong" — that is YOUR job
- Out-of-scope judgment: Without explicit policy, Claude will test everything
- Long context retention: After ~100k tokens, context degrades
Workflow
Phase 1: Per-Target Context Management
"For every target, create a separate folder. Put a .claudemd file in it with the program's policy and scope." — Episode 166 [51:20]
Directory structure:
programs/
├── example-corp/
│ ├── .claudemd # ← THIS IS THE KEY FILE
│ ├── notes/
│ ├── leads/
│ ├── findings/
│ └── scripts/
├── another-target/
│ ├── .claudemd
│ └── ...
.claudemd template:
# Target: Example Corp Bug Bounty Program
## Program URL
https://hackerone.com/example-corp
## Scope — IN
- *.example.com
- api.example.com
- app.example.com (authenticated testing allowed)
- mobile-api.example.com
## Scope — OUT (DO NOT TEST)
- blog.example.com (third-party WordPress)
- status.example.com (StatusPage hosted)
- *.example.dev (staging — explicitly excluded)
- Any domain not listed above
## Rules of Engagement
- NO denial of service testing
- NO social engineering of employees
- NO accessing other users' data beyond proof of concept (read 1 record, stop)
- Rate limit: Max 10 requests/second
- Report vulnerabilities within 24 hours of confirmation
## Authentication
- Test account 1: [email protected] / [use env var AUTH_TOKEN_1]
- Test account 2: [email protected] / [use env var AUTH_TOKEN_2]
- API Key: [use env var EXAMPLE_CORP_API_KEY]
## Tech Stack (Known)
- Frontend: React 18 + Next.js
- Backend: Node.js + Express
- Database: PostgreSQL (inferred from error messages)
- CDN: Cloudflare
- Auth: OAuth 2.0 + JWT
## Priority Targets
1. `/api/v2/users/*` — IDOR testing
2. `/api/v2/billing/*` — Payment logic flaws
3. `/upload/*` — File upload vulnerabilities
4. `/auth/*` — Authentication bypass
## Previous Findings (to avoid duplicates)
- XSS in /search — reported 2024-12, resolved
- IDOR in /api/v1/users/{id} — reported 2025-01, resolved (v2 untested)
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 · 339 lines · 58 tokens per session scan A 762fd566ea0e
ai-pair-hunting-with-claude is a skill published in the GitHub repository ShulkwiSEC/bb-huge (22 stars, last pushed 2mo ago), licensed MIT. It adds 58 tokens to every session and 3,190 once invoked, about $0.0003 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 skills, from other repositories
ai-pair-hunting-with-claude
Configure Claude as a "Pair Hunter" — autonomous overnight hacking, context management via per-target .claudemd files, sub-agent compaction avoidance, and scope enforcement. Based on Critical Thinking Bug Bounty Podcast Episode 166.
strategic-compaction
Detects logical breakpoints in conversations where context compaction makes sense. Fires automatically at session end to evaluate context state. Triggers include: research milestone completion ("ok", "clear", "got it"), failed approaches (3+ attempts), complex changes (5+ files), architecture decisions, and context…
sparc-methodology
SPARC (Specification, Pseudocode, Architecture, Refinement, Completion) comprehensive development methodology with multi-agent orchestration.
claude-skills-for-bug-bounty
Build production-grade Claude Code CLI skills with fallback architecture, TypeScript implementation, and creativity directives. Based on Critical Thinking Bug Bounty Podcast Episode 166 — "Building Claude Skills as a Bug Bounty Hunter.".
skill-context-detection
Auto-detect work context (Dev vs Knowledge) — use to tailor workflows based on current task type.
skill-copilot-provider
GitHub Copilot CLI as optional zero-cost provider via copilot -p programmatic mode.