learning-medusa
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a tutoring protocol where the agent is explicitly directed to request and review code implementations from the user (e.g., in SKILL.md and checkpoint files). This creates an attack surface for indirect prompt injection if the user-provided code contains malicious instructions.
- Ingestion points: Multiple instructions in SKILL.md (Tutoring Protocol Step 5) and checkpoint files (such as checkpoint-module.md) guide the agent to ask the user to share specific code files for review.
- Boundary markers: The skill does not provide instructions to the agent on how to differentiate between code logic and potential embedded instructions, nor does it advise ignoring natural language commands within user-provided snippets.
- Capability inventory: The agent is equipped with the ask_medusa_question tool via the MedusaDocs MCP server and provides instructions for the user to execute shell commands and API requests.
- Sanitization: There are no protocols defined for validating or sanitizing the content of the user-shared code implementations.
Audit Metadata