idea-refine
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from both user input and local project files using tools like
ReadandGrep. It possesses the capability to write files to the local filesystem (e.g.,docs/ideas/). While the skill requires user confirmation before saving artifacts, the prompt instructions lack explicit boundary markers or sanitization logic for the ingested content, representing a surface for indirect prompt injection from the local environment. - [COMMAND_EXECUTION]: The skill contains a shell script (
scripts/idea-refine.sh) designed to initialize documentation directories. The script is restricted to local directory creation (mkdir -p) and does not perform network operations or remote code execution.
Audit Metadata