Skill Risk Checker: Evaluate Agent Permissions Before You Enable

Use this checklist to assess the risk level of any agent skill before enabling it. Covers permission scope, data access, and blast radius.

Skill Risk Checker

What this tool does: Helps you evaluate the risk of enabling any agent skill before you grant permissions.

Quick Risk Assessment

Step 1: Identify the Permission Type

Check what category of permissions the skill requests:

Step 2: Assess the Scope

How much can the skill access?

Step 3: Check Reversibility

Can you undo what the skill does?

Risk Score Calculator

Risk Score = Permission Risk × Scope Multiplier × Reversibility Factor

Low Risk: 1-10
Medium Risk: 11-30
High Risk: 31-100
Critical Risk: 100+

Example Assessment

Skill: "Auto-reply to emails"

1. Permission Type: Send (🔴 High = 8)
2. Scope: All emails (× 10)
3. Reversibility: Irreversible (× 3)

Risk Score: 8 × 10 × 3 = 240 (Critical)

Recommendation: Require approval for each reply, or limit to specific senders.

Detailed Checklist

Data Access Questions

• What specific data can this skill read?
• Is any sensitive data included (passwords, keys, personal info)?
• Can it access more data than needed for the task?
• Where is the data sent (local only, third-party service)?

Action Questions

• What actions can this skill take?
• Are any actions irreversible?
• Could mistakes affect other people?
• Is there a "blast radius" if something goes wrong?

Integration Questions

• What external services does this connect to?
• What credentials are required?
• Can those credentials be scoped down?
• What happens if those credentials leak?

Trust Questions

• Who made this skill?
• Is the source code available for review?
• Are there reviews or security audits?
• How is the skill updated?

Mitigation Strategies

For Medium-Risk Skills

✅ Enable with monitoring ✅ Set up alerts for unusual activity ✅ Review logs regularly ✅ Use sandbox/test data first

For High-Risk Skills

✅ Require approval for each action ✅ Limit to specific use cases ✅ Set strict rate limits ✅ Enable audit logging

For Critical-Risk Skills

✅ Avoid if possible ✅ If necessary, use with human-in-the-loop ✅ Implement multiple approval gates ✅ Regular security reviews

Decision Framework

┌─────────────────────────────────┐
│  Is the benefit worth the risk? │
└─────────────────────────────────┘
         │
         ▼
    ┌────────┐
    │  No    │ → Don't enable
    └────────┘
         │
         ▼ Yes
    ┌────────────────────────┐
    │ Can you reduce scope?  │
    └────────────────────────┘
         │
         ▼ Yes
    ┌────────────────────────┐
    │ Apply least privilege  │
    └────────────────────────┘
         │
         ▼
    ┌────────────────────────┐
    │ Add approval gates for │
    │ irreversible actions   │
    └────────────────────────┘
         │
         ▼
    ┌────────────────────────┐
    │ Enable with monitoring │
    └────────────────────────┘

Related Pages

Hubs

Tools

Glossary

More Resources