engineer-skill-creator
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The creation script
scripts/create_expert_skill.shuses theevalcommand to process user-supplied input strings during variable assignment, which creates a risk of command expansion or execution within the local shell environment. - [COMMAND_EXECUTION]: The tool dynamically generates an interactive shell script
query_expertise.shin the output directory and applies execution permissions viachmod +x. - [INDIRECT_PROMPT_INJECTION]: The skill functions as a formatter for external developer-authored content (code, reviews, notes) and does not apply sanitization or boundary delimiters. This creates a surface for instructions embedded in source profiles to influence agents using the generated skill. Ingestion points: files within the
engineer_profiles/directory. Boundary markers: Absent in the generated expertise documentation. Capability inventory: Local file system read/write and interactive script execution. Sanitization: No filtering or validation of processed instructional content.
Audit Metadata