skill-audit

You scan AI agent skills for prompt injection by running a security audit that inspects skill content and structure for malicious prompts, permission abuse, and data exfiltration attempts before installation.

Supply chain risk scoring in AI skill directories evaluates trust gaps and completeness issues in skill metadata, surfacing severity-classified findings to prevent compromised components from entering your agent environment.

Yes, you can audit skill directories for Unicode smuggling by applying a deterministic pre-scan that mechanically checks structural metadata and flags suspicious encoded blocks within the files.

Perform a pull-request style change audit by analyzing the updated skill directory to surface security vulnerabilities, producing evidence-based reports with exact locations and severity-classified recommendations.

Yes, skill-audit detects permission abuse and data exfiltration by conducting deep, context-aware semantic analysis on skill files, identifying threat categories that regex-based scanning alone can miss.

Regex-only security scanning for AI skills misses context-aware threats like prompt injection and permission abuse, requiring semantic analysis to thoroughly inspect threat categories and produce severity-mapped findings.