construct-causal-model
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is entirely instructional and contains no executable code (Python, Node.js, Shell), network operations, or hardcoded credentials. It defines a conceptual workflow for causal modeling using existing agent capabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data (scientific records and evidence).
- Ingestion points: Processes
variable_records,mechanism_candidates, andevidence_recordsdefined in the input contract in SKILL.md. - Boundary markers: None explicitly defined in the instructions; the skill relies on the structure of the input scientific objects.
- Capability inventory: No risky capabilities such as file writing, network access, or command execution are present in the skill instructions.
- Sanitization: No specific sanitization or validation logic is described for the external content.
Audit Metadata