Offline, evidence-driven security auditing for Agent execution traces.
mumuchongchongchong/security-skills
Skill Codex
Offline reconstruction and security audit of agent execution traces from JSON, JSONL, or pasted logs. Use when Codex needs to review user input, model planning, tool calls, tool results, approvals, and final answers for prompt injection, unauthorized tools, secret exposure, missing approval, repeated-call loops, or…
mumuchongchongchong/security-skills
Skill Codex
Audit third-party Agent and Codex Skills before installation or use. Use for installation checks, Skill poisoning analysis, and supply-chain risk analysis that must inspect scripts, references, resources, trigger descriptions, and declared-versus-observed capabilities, then produce an evidence-driven installation…
mumuchongchongchong/security-skills
Skill Codex
Analyze encoded security logs, HTTP parameters, JSON/JSONL fields, and suspicious text using Base64, Base64URL, URL, Hex, JWT, escape decoding, HTML entities, and bounded Gzip/Zlib decompression. Use for SOC alert triage, preliminary malicious-content analysis, and incident response when offline static inspection is…