---
name: skill-install-trust-verifier
description: OpenClaw skill for governed agent workflows.
---
skill-install-trust-verifier
**Version:** 1.0.0
**Category:** security
**Author:** Iceman / GetAgentIQ.ai
**Build date:** 2026-05-27
What It Does
Performs a pre-install trust check for OpenClaw marketplace skills. Scans skill files for risk patterns (shell execution, network calls, credential reads, prompt injection, config/cron mutation) and scores trust signals (SKILL.md present, self-test documented, safety note, restore point reference, stdlib-only).
Returns INSTALL_OK / REVIEW_BEFORE_INSTALL / DO_NOT_INSTALL recommendation.
Who It's For
→ OpenClaw users who want to vet marketplace skills before installation (addresses documented community concern about junk/malicious uploads)
→ Security-conscious operators running managed OpenClaw deployments
→ Teams with strict change-control requirements before adding new skills
Usage
```bash
python3 scripts/skill_install_trust_verifier.py --target /path/to/skill-directory --json
python3 scripts/skill_install_trust_verifier.py --self-test --json
```
Output
JSON fields: `ok`, `recommendation`, `risk_score`, `trust_score`, `trust_signals`, `files_scanned`, `findings`, `next_steps`.
QA Retest Fixture
```bash
python3 -m py_compile scripts/skill_install_trust_verifier.py
python3 scripts/skill_install_trust_verifier.py --self-test --json
python3 scripts/skill_install_trust_verifier.py --target skills/upgrade-risk-preflight-checker --json
```
Expected: self-test `ok=true`; clean skill target `recommendation=INSTALL_OK`.
Safety
Read-only, stdlib-only, no network, no subprocess, no shell execution, no secret values printed in findings. Nosec annotations on detector-definition lines only; runtime scanner behavior preserved.
Restore point
Create a restore point before installing in any production workspace.
Release evidence
Reviewed in the 2026-06-15 GetAgentIQ warehouse propagation cycle.