let-it-rip
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is configured to follow instructions from external documents (markdown files in specific project directories), which introduces a surface for indirect prompt injection.
- Ingestion points: Instructions are read from
docs/experiments-doc/<related_name.md>anddocs/experiments-pipeline/<related_name.md>. - Boundary markers: Absent; there are no clear delimiters or system instructions used to prevent the agent from obeying malicious directives embedded within the documentation files.
- Capability inventory: The sub-agent has the capability to create/modify files, run notebook code cells, execute git commits, and invoke additional tools like
exp-review. - Sanitization: Absent; the content of the external documentation is not validated or sanitized before being treated as a task description for the agent.
Audit Metadata