bmad-agent-creative-muse
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust state management system using a local 'sanctum' directory. It includes clear instructions to avoid sensitive data such as .env files, credentials, and secrets, ensuring that the agent's operations remain within safe boundaries.
- [COMMAND_EXECUTION]: Includes a Python initialization script ('init-sanctum.py') that automates the setup of the agent's environment. This script performs deterministic file system operations like directory creation and template-based file copying. Additionally, the skill provides a framework for 'Capability Authoring' where the agent can generate and execute Python scripts locally within its sanctum to extend its functionality.
- [PROMPT_INJECTION]: The 'Pulse' mechanism involves reading raw session logs to distill insights into long-term memory. This represents a surface for indirect prompt injection. Ingestion points: sessions/YYYY-MM-DD.md. Boundary markers: Memory philosophy defined in memory-guidance.md. Capability inventory: Local file system access and capability-based script execution. Sanitization: Handled through agent-led curation logic.
- [DATA_EXFILTRATION]: No evidence of unauthorized data transfer was found. The agent's operations are strictly limited to the local project directory and its own memory files, with no network-enabled commands or tools detected.
Audit Metadata