flag-cleanup
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses a bundled script
${CLAUDE_PLUGIN_ROOT}/scripts/gb-callfor all API communications. This local execution pattern is standard for modular skills.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes external data from GrowthBook API endpoints.\n - Ingestion points: Data enters the context from
/api/v2/features,/api/v1/code-refs, and/api/v1/experiments(SKILL.md).\n - Boundary markers: No specific delimiters are used to wrap the API-provided content.\n
- Capability inventory: Employs the
Bashtool (forgb-call) and has the ability to read/edit the local codebase (SKILL.md).\n - Sanitization: API content is used directly without explicit sanitization or validation against embedded instructions.\n- [SAFE]: Credentials for the GrowthBook API are managed through environment variables or a standard local configuration file (
~/.config/growthbook/.env). This is noted as a safe practice for credential storage.
Audit Metadata