ClawHub
Back to skill

Security audit

checklist

Security checks for vulnerabilities and agentic risk

Overview

This skill helps users design operational checklists without requesting sensitive access or autonomous high-impact actions.

Use this as a checklist-design aid. Validate checklist items with the people responsible for the operation, especially in medical, financial, or deployment workflows, and retain explicit approval for irreversible actions.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · examples/ai-agent-preflight-checklists-2023-2026.md (reported line 23)May include surrounding context.

md
probe. Write the 8-item DO-CONFIRM gate now — evals-on-this-version, injection probe, least-privilege tools, spend cap, rollback, human gate on irreversible actions, PII check, incident owner — and make the promoting engineer own it. It costs 10 minutes a ship and is the cheapest "enterprise-ready" signal you can build.

**Founder takeaway (SMB):** Same discipline, smaller surface: before turning on any AI tool that talks to customers or touches money, a 6-item READ-DO — what can it send/spend without asking me? · did I test the 5 questions customers actually ask? · is there a human handoff? · a daily cost cap? · do I know how to turn it off? · who checks its output the first week? A restaurant's reservation bot and a Fortune-500 agent fail the same way: a known step, skipped.

*Sources: NIST (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0), January 2023. https://www.nist.gov/itl/ai-risk-management-framework — the institutional codification of pre-deployment AI c

Static analysis

No suspicious patterns detected.