Skill Security Audit

This TIKAZ Edition is designed, integrated, refactored, and continuously maintained by TIKAZ as a suite safety gate. External scanners remain separately attributed when actually used.

Inputs and workflow

Inspect SKILL.md, scripts, manifests, dependencies, assets, links, and install commands. Classify findings as blocker, high, medium, or informational and cite exact files and lines. Check for secret collection, browser/session access, recursive deletion, privilege escalation, remote execution, obfuscated payloads, unpinned dependencies, and instructions that override user intent.

Output contract

Return scope, inventory, findings with evidence, dependency and permission notes, unresolved dynamic behavior, and one recommendation: approve, approve-with-conditions, or reject.

Validation and fallback

Inspect unpacked source rather than trusting the README. If archives, binaries, generated payloads, or network behavior cannot be inspected, mark them unresolved and avoid approval. Static approval is not a guarantee of safety; use least privilege during installation.

Example and limits

Use skill-security-audit on this Skill directory before installation. Classify blockers, cite exact evidence, and state any behavior static inspection cannot verify.

The audit does not execute suspicious code or prove runtime safety.