audit-skill

Pass

Audited by Gen Agent Trust Hub on May 26, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for cloning remote repositories using git clone to a temporary directory for auditing purposes.
  • [REMOTE_CODE_EXECUTION]: Instructs the agent to run the cyber-skills npm package via npx to perform automated security and structure validation. This is a functional requirement for the skill's auditing purpose.
  • [COMMAND_EXECUTION]: Mentions high-risk commands (e.g., rm -rf, :(){ :|:& };:) as part of its security check definitions (E1). These are reference patterns for the auditing process and not directives for execution.
  • [PROMPT_INJECTION]: Ingests untrusted external SKILL.md files for analysis, which creates a surface for indirect prompt injection. The skill addresses this with a dedicated 'Sandboxing' section that instructs the agent to treat all audited content as data and explicitly forbids acting on any instructions found within those files. It also includes common injection strings as data patterns for the auditor to identify.
Audit Metadata
Risk Level
SAFE
Analyzed
May 26, 2026, 12:25 AM
Security Audit — agent-trust-hub — audit-skill