# Safe Agent Skills vs Agent Skill Safety Kit

**Short version:** Safe Agent Skills is useful for catalog-scale discovery and risk browsing. Agent Skill Safety Kit is the local review workflow for deciding whether a specific skill should be installed, listed, sold, delegated, sandboxed, or rejected in your environment.

URL: https://jarvislandingdeploy.vercel.app/safe-agent-skills-vs-agent-skill-safety-kit.html

Last updated: 2026-09-25.

## Where each layer fits

| Question | Safe Agent Skills | Agent Skill Safety Kit |
|---|---|---|
| Which public skills look risky at a glance? | Strong fit: catalog, risk labels, and safety data sheets. | Focused on reviewing a chosen skill, not replacing a broad catalog. |
| What authority will this skill receive inside our agent? | Catalog risk informs the decision, but cannot know local tools, secrets, workflows, customers, memory stores, or approval rules. | Primary fit: permission manifest defines allowed reads, writes, browser/API actions, memory updates, messages, spend, and approvals. |
| Can we explain why we approved or rejected the skill? | Useful evidence, but final rationale needs local policy and disposition. | Primary fit: produces install / install with limits / ask for changes / reject / manual review verdict. |
| What proof should the skill leave after it acts? | Catalog risk labels usually stop before runtime receipt design. | Primary fit: receipt fields, before/after state, source-of-truth checks, exception behavior, and rollback notes. |

## Practical workflow

1. Discover candidate skills and obvious red flags in Safe Agent Skills or a similar catalog.
2. Pin provenance: source URL, version, author/vendor identity, package hash or commit, and claimed use case.
3. Scan and inspect the package; record findings as evidence, not the whole decision.
4. Write a permission manifest covering reads, writes, tools, memory, messages, outbound actions, spend, and human gates.
5. Score identity, inputs, data access, tool authority, output contract, receipts, memory, exceptions, evals, rollback, and portability.
6. Decide the launch verdict: install, install with limits, sandbox, ask for changes, reject, or request a short custom review.

## Free artifacts

- Agent Skill Safety Kit: https://jarvislandingdeploy.vercel.app/agent-skill-safety-kit.html
- Scorecard: https://jarvislandingdeploy.vercel.app/agent-skill-safety-scorecard.html
- Permission manifest template: https://jarvislandingdeploy.vercel.app/agent-skill-permission-manifest-template.html
- Permission manifest validator: https://jarvislandingdeploy.vercel.app/agent-skill-permission-manifest-validator.html
- Sample mini-report: https://jarvislandingdeploy.vercel.app/agent-skill-safety-sample-report.html

Use the $29 Custom Agent Audit Mini-Report only for high-authority skills where the free review exposes meaningful gaps: https://jarvislandingdeploy.vercel.app/products/custom-agent-audit-mini-report.html