Systematically verifies suspected security bugs to eliminate false positives. Produces TRUE POSITIVE or FALSE POSITIVE verdicts with documented evidence for each bug.
Build, debug, and optimize Claude API / Anthropic SDK apps. Apps built with this skill should include prompt caching. Also handles migrating existing Claude API code between Claude model versions (4.5 → 4.6, 4.6 → 4.7, retired-model replacements). TRIGGER when: code imports `anthropic`/`@anthropic-ai/sdk`; user asks for the Claude API, Anthropic SDK, or Managed Agents; user adds/modifies/tunes a Claude feature (caching, thinking, compaction, tool use, batch, files, citations, memory) or model (Opus/Sonnet/Haiku) in a file; questions about prompt caching / cache hit rate in an Anthropic SDK project. SKIP: file imports `openai`/other-provider SDK, filename like `*-openai.py`/`*-generic.py`, provider-neutral code, general programming/ML.
Use when auditing a web app for security vulnerabilities. Triggers on: 'audit my project', 'how could this get hacked', 'find bugs in this codebase', 'security review', 'pentest', 'bug bounty scope for X', 'GHSA collaborator invite', 'security audit this'. Handles both internal audits (sitting in your own codebase, filesystem access) and external audits (only URL + written authorization). Contains: context detection (inside-vs-outside), a source-code audit checklist (auth bugs, IDOR, SSRF, injection, crypto misuse, file handling, rate-limits, Docker config), a black-box probing checklist (subdomain enum, port scan, TLS/DNS, admin-panel leakage, error-shape oracles, version fingerprint), a server/infra sweep (management ports, cloud metadata, CDN bypass, backup files, container image CVEs), a rate-limit deep-dive (distributed-counter vs in-memory, XFF trust, cost-inflation DoS), and — after you have findings — advisory writing via gh api and patch delivery when the temporary private fork is gated to the mainta