Cloudflare's security-audit-skill runs six-phase agent audits with independent verification
alex_verem · x · 2026-10-22
Cloudflare's MIT-licensed security-audit-skill (38.5k stars) is the single-repo origin of its fleet-wide vulnerability discovery harness. It turns a coding agent into a structured security auditor across six phases:
- Reconnaissance — map architecture, trust boundaries, input surfaces into architecture.md and coverage-ledger.json
- Coverage-led hunting — assign isolated hunter agents from ledger units, with coverage critics finding gaps
- Candidate validation — a fresh verifier tries to disprove every unique candidate
- Structured output — confirmed/needsvalidation/rejected records in findings.json validated against report-schema.json
- Independent record verification — fresh agents verify final claims; material replacements get another verifier
- Target-neutral reporting — generates REPORT.md, FINDINGS-DETAIL.md and NEEDS-VALIDATION.md
More from coding & agent
- Hooking an AI agent to a local Obsidian vault on a Mac mini: what worked and what didn't — techlatest_net · 2026-10-23
- How much memory should an AI agent really have? Reddit debates the tradeoffs — Mission_Plankton_115 · 2026-10-23
- Benchhand: open-source Rust MCP client with editable tool approvals and per-agent models — AnshSuc · 2026-10-23
- CrawlForge MCP ships 24 releases in 24 days: free failed calls, 1-credit repeat reads — slymes-20 · 2026-10-23
- Hands-on Orq AI review: LLM gateway, evals and guardrails in one EU-hosted platform — FigureReasonable4468 · 2026-10-23
- I spent $200 benchmarking AI bug-fixing: Opus tops at 94%, DeepSeek matches for 1/26 the cost — narrow-adventure · 2026-10-23