darwin
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill 'darwin' performs high-level ecosystem auditing and lifecycle detection. It calculates health scores (EFS) and relevance scores (RS) based on verifiable signals without assuming elevated privileges or performing dangerous operations. Its logic is transparent and focused on project maintenance.
- [EXTERNAL_DOWNLOADS]: The skill documentation references external research, benchmarks, and technical blogs from well-known and trusted organizations including Anthropic, Microsoft, and LangChain. These references serve as informational context for the skill's fitness criteria and do not involve the download or execution of untrusted scripts.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from untrusted sources such as git commit messages, project file content, and cross-agent journals to perform its analysis. While this presents an indirect prompt injection surface, the skill implements a 'Sense-Assess-Verify' workflow with confidence scoring to mitigate the risk of obeying embedded instructions in processed data.
- [COMMAND_EXECUTION]: The instructions involve the use of read-only git commands (e.g.,
git log,git shortlog,git tag) and filesystem searches (e.g.,find) to collect signals for project phase detection. These operations are restricted to signal collection and do not expose the system to arbitrary command injection.
Audit Metadata