idea-incubation
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled script (
node scripts/validate-workspace.js) to verify the integrity of the workspace after state changes. The execution uses a static path and operates on agent-managed directories, minimizing risk. - [PROMPT_INJECTION]: The skill processes user input and external research data, which presents a surface for indirect prompt injection. This is mitigated by the skill's rigid 12-phase state machine and specific output quality gates.
- Ingestion points: Raw user idea inputs and web search results stored in research artifacts.
- Boundary markers: The skill uses structured Markdown and YAML sections for data storage but does not explicitly use delimiters to sequester untrusted external content.
- Capability inventory: Includes file system write access, web search, and local script execution.
- Sanitization: The workflow relies on 'quality gates' to verify content, though no explicit string sanitization for research data is described.
Audit Metadata