principle-attack-the-premise
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The analyzed skill files (
SKILL.mdandagents/openai.yaml) contain only documentation and configuration. No security risks were identified. - Instructional Content: The body of
SKILL.mddescribes a logical framework for debugging ('Attack the Premise'). It does not contain prompt injection patterns or commands. - Metadata and Configuration: YAML frontmatter and the agent configuration file use restrictive settings (e.g.,
disable-model-invocation: true,allow_implicit_invocation: false) which limit the skill's capabilities. - Relative References: The skill references other principle documents using relative paths, which is standard documentation practice and does not involve external downloads or unverifiable dependencies.
Audit Metadata