idea-darwin
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted local files (
ideas.mdandstimuli.md) to perform its core functions of 'evolution' and 'mutation'. - Ingestion points: Content is read from
ideas.mdandstimuli.mdto be scored and evolved. - Boundary markers: The skill does not specify the use of delimiters or 'ignore embedded instructions' warnings when processing these files.
- Capability inventory: The skill involves text analysis, logical refinement, and creative synthesis based on the input data.
- Sanitization: There are no instructions for sanitizing or validating the content of the input files before they are processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill's documentation references an external, untrusted source for the project's code and installation.
- Evidence: References a GitHub repository (
github.com/warmskull/idea-darwin) and provides an installation command (clawhub install idea-darwin) for a third-party platform that is not recognized as a trusted vendor.
Audit Metadata