Claude-BugHunter is a self-contained skill bundle for the Claude Code skills system that transforms the agent into a senior bug-hunting researcher or red-team operator through 83 auto-loading skills and 15 slash commands. Built by Sachin Sharma, the bundle draws on 681 disclosed HackerOne reports across 24 vulnerability classes, with 433 patterns individually cited and auditable. It stacks four layers: a 5-phase non-linear workflow and critical-thinking framework (bb-methodology, redteam-mindset); 58 hunt-* webapp skills covering detection patterns, payloads, bypass tables, and chain templates for classes like XSS, SQLi, SSRF, IDOR, and SSTI; 10 enterprise platform attack skills targeting M365/Entra ID, Okta, vCenter, SSL-VPN appliances (Cisco, Fortinet, Citrix, Palo Alto, Pulse, SonicWall, F5), SharePoint, and cloud IAM with 2024–2026 CVE chains and post-credential escalation; and six reporting/validation skills implementing a 7-Question Gate, VRT-aware severity, OOS rebuttals, PII redaction, and red-team deliverables. Skills load automatically by topic — no manual invocation — and install via a namespaced plugin (Option A) or a copy install that includes the cbh CLI and engagement scaffolder (Option B). The --all flag ports skills to five harnesses: Claude Code, OpenCode, OpenAI Codex CLI, Hermes Agent, and Google AntiGravity, while --burp-mcp wires the Burp MCP server. The bundle explicitly scopes to external attack surfaces — bug bounty, webapp pentesting, and external red-team initial access — excluding internal AD post-exploitation, C2 frameworks, and evasion. It enforces authorized-use posture with scope validation gates and recommends Anthropic's Cyber Verification Program for safeguard adjustment. Dual-licensed: MIT for code, CC BY 4.0 for content.

Read original