
3.5K-star offensive-security skill — 9 AI agents, 35 tools, zero-subscription path via Ollama.
shuvonsec/claude-bug-bounty), now at 3,500 stars and 613 forks, and the first offensive-security Agent Skill covered in this channel. 1"Can an attacker do this RIGHT NOW against a real user who has taken NO unusual actions — and does it cause real harm (stolen money, leaked PII, account takeover, code execution)? If the answer is NO — STOP. Do not write. Do not explore further. Move on."
/pickup target.com.git clone https://github.com/shuvonsec/claude-bug-bounty
cd claude-bug-bounty
./install.sh --agent standalone
bughunter setupbughunter setup runs an interactive provider selector. The fully free, offline path: choose Ollama, then pull qwen2.5:14b (~9 GB). Groq's free API tier is the fastest cloud option. DeepSeek costs roughly $0.001 per 1K tokens../install_tools.shnpx skills add shuvonsec/claude-bug-bountySKILL.md into your project, then open the directory in Claude Code. The slash commands (/recon, /hunt, /validate, /report, /autopilot) become available immediately.AGENTS.md for per-framework config. 3bughunter recon target.com # passive + active recon, subdomain enum
bughunter hunt target.com # vulnerability discovery across all modules
bughunter validate "finding" # 7-Question Gate — pass/fail before writing
bughunter report # generates platform-ready submissionbughunter autopilot target.com--paranoid flag for /autopilot in Claude Code that increases validation strictness.
/web3-audit contract.sol # Claude Code slash command
/token-scan <contract-address> # Solana and EVM supported/scope <asset> — verifies an asset is in-scope before you touch it/triage — runs a 2-minute pass/fail check before deeper analysis/pickup target.com — resumes from the last session checkpoint/bypass-403 — dedicated forbidden-response bypass module/secrets-hunt — credential and secret discovery across the target surfacebughunter chatdemo/ directory with a locally-served vulnerable target (python3 serve.py) for trying the workflow without hitting a live program.| Issue | Details | Status |
|---|---|---|
| Grok API errors | Issue #80 (filed Jun 18): xAI/Grok API returns "400 Client Error: Bad Request" on chat calls | Open 10 |
| No Windows native support | macOS and Linux only; Windows requires WSL | By design 1 |
| External tool dependency | Full capability requires 35 tools from install_tools.sh; partial installs degrade scan coverage silently | Known limitation 1 |
| Rate limits and WAF | High-volume scan modes can trigger target WAFs; built-in circuit breakers exist but require user awareness | Known limitation 5 |
| macOS bash 3.2 | Auth crash on macOS bash 3.2 — fixed in v4.3.1 | Resolved 11 |
| PatternDB performance | O(n²) slowdown at 10K entries (5+ min) — fixed in v4.2.1, now ~2 sec | Resolved 11 |
/recon→/validate→/report as native slash commands| Repository | shuvonsec/claude-bug-bounty 1 |
| License | MIT 1 |
| Latest version | v5.0.0 (released Jun 9, 2026) — reduced false positives + repo optimization 11 |
| Stars / forks / commits | 3,500 stars · 613 forks · 140 commits 1 |
| Supported agents | Claude Code, Codex, OpenCode, Pi Agent (standalone CLI runs in any terminal) 3 |
| AI providers | Ollama (local, free) · Groq (free tier) · DeepSeek (~$0.001/1K tok) · Claude API · OpenAI 1 |
| Prerequisites | macOS or Linux, Go, Python 3.9+, jq; Windows via WSL 1 |
| Coverage | 20 Web2 vulnerability classes · 10 Web3/smart contract categories 1 |
| Platforms | HackerOne, Bugcrowd, Intigriti, Immunefi 1 |
| Author | Shuvon Md Shariar Shanaz (@shuvonsec) 1 |
| Open issues | #80 — Grok API 400 error (Jun 18, 2026) 10 |
github.com
raw.githubusercontent.com
raw.githubusercontent.com
raw.githubusercontent.com
raw.githubusercontent.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.