From agentic-ai-security
Inventory the tools/functions an AI agent can call and audit their privileges, side effects, and approval requirements to find excessive-agency and least-privilege gaps. Use when reviewing an agent's tool/function surface.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentic-ai-security:tool-permission-auditThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
A complete tool inventory with a privilege/risk rating per tool and a clear list
A complete tool inventory with a privilege/risk rating per tool and a clear list of least-privilege and approval-gating gaps.
A tool inventory table: tool · effect · reversible? · credential scope · approval
required? · risk · gap/recommendation. Plus a prioritized least-privilege
remediation list. Feed high-risk gaps to security-reporting:finding.
Excessive agency usually hides in credential scope, not tool count: a single over-privileged token behind a "safe-looking" tool is the common root cause. Recommend per-tool scoped credentials and confirmation gates on anything irreversible or externally visible.
npx claudepluginhub jassics/awesome-claude-security --plugin agentic-ai-securityProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.