devtu-self-evolve
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONSAFE
Full Analysis
- [DYNAMIC_EXECUTION]: The skill manages a lifecycle that includes generating and modifying Python tool code (via dispatches to
devtu-create-toolanddevtu-fix-tool) and executing it usingpython -m tooluniverse.cli. It also performs syntax verification usingpython -con dynamic modules. - [INDIRECT_PROMPT_INJECTION]: The workflow relies on feedback from autonomous 'researcher persona' agents to identify and implement bug fixes. This creates a surface where subagent output can influence the code written to the repository. The skill mitigates this by requiring manual/CLI verification of all agent-reported issues before implementation.
- [COMMAND_EXECUTION]: The skill makes extensive use of system CLI tools, including
git(fetch, rebase, push, stash) andgh(pr list, pr create, pr view), to manage the development workflow on themims-harvard/ToolUniverserepository. - [SAFE]: The external resources referenced, including the GitHub repository and various life-science APIs (such as CIViC, KEGG, ClinVar, and CPIC), are well-known services or belong to the skill author's infrastructure, consistent with the stated research and development purpose.
Audit Metadata