write-launch
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local TypeScript scripts provided in the package, such as 'manifest-sync.ts' and 'query-performance.ts'. These scripts are used to synchronize the project's artifact manifest and retrieve performance data from local TSV files. These operations are localized to the project directory and serve the skill's primary functional purpose.
- [PROMPT_INJECTION]: The skill's architecture involves ingesting data from project files (e.g., icp-research.md, brand/BRAND.md) and user-provided 'Cold Start' responses, which are then interpolated into prompts for sub-agents. This constitutes an indirect prompt injection surface. However, the skill implements a multi-agent validation process where a 'Guard Checker Agent' specifically re-scans output for violations like vote solicitation or undisclosed promotions, and a 'Critic Agent' scores the result against a 5-dimension rubric to ensure quality and safety.
Audit Metadata