idea-refine
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a shell script (scripts/idea-refine.sh) intended to initialize the project structure by creating a directory (docs/ideas). The script is minimal, performs no network operations, and uses standard directory creation commands.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from both user arguments ($ARGUMENTS) and local project files using tools like Glob, Grep, and Read. This creates a vulnerability surface where instructions embedded in project files or user input could potentially influence agent behavior.
- Ingestion points: User-provided concepts via $ARGUMENTS and local file contents retrieved via Glob, Grep, and Read tools in SKILL.md.
- Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore potentially malicious commands embedded within the data it scans.
- Capability inventory: The agent has the ability to execute a local shell script, read arbitrary files within the project scope, and write refined ideas to the docs/ideas/ directory.
- Sanitization: No explicit sanitization or filtering of external content is specified before the data is processed by the agent.
Audit Metadata