lw-site-manager-extend-abilities

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents a pattern for building custom tools ('abilities') that accept input from AI agents, which introduces a surface for indirect prompt injection attacks.
  • Ingestion points: The input_schema definitions in the NewsletterAbilitiesRegistrar examples in SKILL.md represent the entry point for untrusted agent data.
  • Boundary markers: The instructional code does not include delimiters or specific instructions to help the model distinguish between data and potential commands within the input.
  • Capability inventory: The execute_callback parameters (e.g., in NewsletterManager) grant the abilities the power to modify site state or perform network operations based on agent input.
  • Sanitization: The service layer examples do not demonstrate input validation or sanitization before processing the arguments received from the agent.
  • [SAFE]: The skill references official documentation and source code repositories for developer context.
  • Evidence: Provides links to the plugin's GitHub repository and the official WordPress developer portal.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:06 PM
Security Audit — agent-trust-hub — lw-site-manager-extend-abilities