create-experiment
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (hypothesis, metrics, treatments, and rules) and passes them to another skill (manage-experiments/SKILL.md) for processing.
- Ingestion points: User-supplied fields like hypothesis and rule processed in SKILL.md.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in this entrypoint.
- Capability inventory: Does not invoke tools directly; all capability execution is delegated to the manage-experiments workflow which interacts with the Harness MCP server.
- Sanitization: No input validation or sanitization logic is performed on the user-provided experiment parameters before delegation.
Audit Metadata