Agent Skill Trust Check

Static pre-install trust check for agent skills, SKILL.md files, and skill marketplace listings.

Agent skill marketplaces are useful, but installing a skill is still a trust decision. This repo provides a small local scanner and a portable SKILL.md review guide for checking public skill text before it gets installed.

Run

git clone https://github.com/TateLyman/agent-skill-trust-check.git
cd agent-skill-trust-check
npm run check
node bin/agent-skill-trust-check.js ./SKILL.md

Run against a public GitHub/raw/Gist URL from the checkout:

node bin/agent-skill-trust-check.js https://raw.githubusercontent.com/owner/repo/main/SKILL.md --json

Run the marketplace-safe stdin wrapper:

node bin/agent-skill-trust-check-stdin.js < ./SKILL.md

Planned npm command after the first npm publish:

npx --yes agent-skill-trust-check ./SKILL.md

What It Checks

• Shell/process execution.
• Destructive command patterns.
• Secret or credential access.
• Wallet/payment/signing behavior.
• Remote network/webhook output.
• Persistence/background behavior.
• Prompt-boundary language.
• Missing provenance signals such as license, tests, permissions, versioning, source, and uninstall notes.

Output

The CLI returns:

• verdict
• risk_score
• findings
• positives
• missing_signals
• patch_order

The stdin wrapper always returns compact JSON and does not read local files, fetch URLs, or spawn subprocesses. It is the safer entry point for agent-skill marketplaces that call tools with stdin/stdout contracts.

Marketplace Review

For a paid marketplace-grade pass, use:

https://orkai.ai/skills/agent-skill-trust-check

For private review work:

https://tateprograms.com/payments.html

For creator marketplace submission copy, see:

marketplace/agensi-listing.md

Additional marketplace-ready skill files:

• marketplace/paperclip-skill.md for Paperclip Skills publish flow.
• marketplace/skillshop-skill.md for SkillShop private-repo listings.

Boundary

This scanner is a static text pass. It does not execute the target skill and it does not replace sandboxing, code review, dependency review, or runtime monitoring.

License

MIT