product-decision-agent
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill incorporates a local utility script (
scripts/quality_gate.py) used to validate the quality and persona consistency of the agent's output. The script performs local regex-based linting and does not engage in network activity, sensitive file access, or privilege escalation. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a reasoning engine for user-provided business problems, which introduces a surface for indirect prompt injection.
- Ingestion points: User-submitted business scenarios, project data, and team conflict descriptions in
SKILL.md. - Boundary markers: The agent is instructed to perform a 'Silent Reasoning' phase to separate 'Hard Facts' from 'Assumptions' and 'Solutions' provided by the user, providing a logical boundary for processing untrusted data.
- Capability inventory: The skill maintains the ability to read internal reference files (
references/) and execute a local Python-based linter for quality assurance. - Sanitization: The
scripts/quality_gate.pyscript serves as a post-processing filter to ensure the agent's output remains within safety and stylistic boundaries, specifically flagging vague advice or methodology leakage.
Audit Metadata