agent-self-assessment
由 roosch269 提供
Web & Frontend Development简介
Security self-assessment tool for AI agents.
描述
该技能的功能
Runs a 10-check security and compliance self-assessment on an AI agent's own configuration. Covers decision boundaries, credential scoping, audit trails, plane separation, and EU AI Act readiness across Articles 6, 14, and 50. Produces a structured RED/AMBER/GREEN rated report across three domains: security, EU AI Act readiness, and data governance.
It runs entirely locally with no data sent externally, covering 10 structured checks across security and regulatory domains that would otherwise require manual review against multiple frameworks.
何时使用
- Auditing an agent's security posture before a production deployment
- Preparing an agent for the August 2026 EU AI Act enforcement deadline
- Checking whether credentials are properly scoped across tools and services
- Generating a compliance report for an internal governance review
- Identifying missing human oversight mechanisms in an autonomous agent
示例工作流
User asks: "Read SKILL.md and run the agent compliance assessment"
- 1
Reads local configuration files and inspects environment structure
- 2
Checks each tool for consequential write, send, or pay operations and whether human gates exist
- 3
Reviews EU AI Act readiness across transparency disclosure, risk classification, and human override capability
- 4
Evaluates data processing practices, retention policies, and erasure capability
- 5
Outputs a structured report with per-check RAG ratings, findings, and prioritized remediation actions
A formatted compliance report with RED/AMBER/GREEN ratings for all 10 checks, an overall security posture label (SECURE / HARDENING NEEDED / CRITICAL), and a prioritized action list
安装
clawhub install roosch269/agent-self-assessment运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。