security-audit

A coding-agent skill for multi-phase security audits. Served from this host; files under /security-audit/.

Upstream corpus (agent-neutral)

• SKILL.md — workflow, evidence bar, execution safety
• report-schema.json — findings schema

DeepSeek Harness build

• dsh/SKILL.md — harness-adapted entry point
• dsh/DSH-ORCHESTRATION.md — subagent / workflow phase mapping

Phases

• RECONNAISSANCE.md — phase 1
• HUNTING.md — phase 2
• VALIDATION-AND-REPORTING.md — phases 3–6

Attack classes

• ATTACK-CLASSES.md — core, wildcard, obvious things
• AI-AND-LLM.md
• CLIENT-SIDE.md
• CLOUD-AND-DEPLOYMENT.md
• DATA-ISOLATION-AND-LIFECYCLE.md
• DESKTOP-MOBILE-AND-LOCAL-IPC.md
• MEMORY-SAFETY-AND-BINARY.md
• PROTOCOLS-RPC-AND-MESSAGING.md
• RESOURCE-EXHAUSTION-AND-AVAILABILITY.md
• SUPPLY-CHAIN-AND-RELEASE.md
• WEB-PROTOCOL-AND-AUTH.md

Validators

• validate-findings.cjs
• validate-coverage-ledger.cjs
• validate-findings.test.cjs
• validate-coverage-ledger.test.cjs

Corpus derived from cloudflare/security-audit-skill (MIT). Validators require POSIX file semantics and do not run on Windows.