skills-qa

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill contains a 'Heuristic trust scan' section specifically designed to detect and flag prompt injection attempts in other skills, such as instructions to ignore guardrails or activate developer mode overrides.
  • [DATA_EXFILTRATION]: The instructions include checks for suspicious data handling in analyzed content, identifying requests for credentials, environment secrets, or unauthorized network operations.
  • [INDIRECT_PROMPT_INJECTION]: The skill performs analysis on external data (the target SKILL.md file). While this represents a potential ingestion point for malicious instructions, the skill implements defensive measures by requiring the agent to quote the raw text rather than relying on memory, and by applying a dedicated security scanning process to that input.
  • [COMMAND_EXECUTION]: The frontmatter explicitly sets 'disable-model-invocation: true', which restricts the skill from being automatically triggered by the model, reducing the risk of unintended execution chains.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 02:12 PM
Security Audit — agent-trust-hub — skills-qa