Refinery Agent Pattern
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional Markdown files, templates, and checklists. It does not include any executable scripts (e.g., Python, JavaScript, Shell) or binaries.
- [SAFE]: No network operations, sensitive file access, or hardcoded credentials were detected. The workflow involves reading from and writing to local project directories (
project-idea/andproject-docs/). - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted input from the
project-idea/directory to generate documentation. While this constitutes an attack surface for indirect prompt injection, the agent's scope is strictly limited to text transformation and documentation generation, presenting no significant security risk. - Ingestion points:
project-idea/index.mdandproject-idea/amendments/*.md(SKILL.md). - Boundary markers: Absent; the agent is instructed to read the raw content directly.
- Capability inventory: The agent performs text extraction, organization, and file generation (Markdown/JSON). No subprocess execution or network capabilities are defined.
- Sanitization: Absent.
Audit Metadata