ralph
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The skill's primary purpose is to ingest untrusted data (a PRD provided by a user) and transform it into a task list (
prd.json) for the 'Ralph' autonomous agent system. - Ingestion points: The skill takes a 'PRD (markdown file or text)' as input (SKILL.md).
- Boundary markers: There are no explicit boundary markers or instructions to the LLM to ignore embedded instructions within the input PRD.
- Capability inventory: While this specific skill only performs a conversion, the resulting
prd.jsonis explicitly designed for 'autonomous execution' (SKILL.md) which includes capabilities like running database migrations, executing shell commands, and interacting with a web browser. - Sanitization: No sanitization or validation of the PRD content is performed; it is mapped directly into JSON fields like
acceptanceCriteriaanddescriptionwhich guide the agent's future actions.
Audit Metadata