Skill Safety Vetting Layer 20260730 - v1.0

What It Does

Skill Safety Vetting Layer 20260730 converts redacted pre-install review notes into a deterministic safety evidence report. It checks source provenance, source digest, version pin, permissions, dependencies, static review, malware scan, network boundary, filesystem boundary, compatibility evidence, sandbox testing, rollback planning, and reviewer signoff.

The skill is advisory and local-only. It does not install code, execute third-party code, alter permissions, call networks, mutate files, or certify safety by itself.

Use Cases

• Marketplace buyers checking whether a skill has enough provenance and permission evidence before install.
• Security reviewers standardizing pre-install PASS, REVIEW, BLOCKED, and ERROR outcomes.
• Operators preparing dependency, boundary, sandbox, and rollback evidence for independent QA.

Requirements

• Python 3 standard library.
• No API keys required.
• OpenClaw skill workspace with the included scripts/skill_safety_vetting_layer.py.
• Redacted local evidence notes supplied with --notes or --text.

Example Usage

python3 scripts/skill_safety_vetting_layer.py --notes redacted-vetting-notes.txt --json

Expected output is structured JSON containing a verdict such as PASS, REVIEW, BLOCKED, or ERROR, plus missing safety evidence, risk signals, and mutated_files: [].

Price

$4.99

Website Card Copy

Review redacted skill safety notes for provenance, digests, permissions, dependencies, static review, malware scan evidence, boundaries, compatibility, sandbox tests, rollback planning, and signoff before install.