review-learning-content
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using
bunx exercode-problem. This involves running an external package to perform judging tasks. - [DYNAMIC_EXECUTION]: The instructions require the agent to execute code provided in the repository, specifically 'model answers, templates, and judges', including custom
judge.tsscripts. Executing code provided in the materials being reviewed is a form of dynamic execution that could be exploited if the source material is malicious. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from learning materials, questions, and problems. This represents an attack surface where malicious instructions could be hidden in the content.
- Ingestion points: Processes learning content, markdown files, embedded questions, and Judge problem definitions.
- Boundary markers: No specific delimiters or safety instructions are used to separate the content from the agent's instructions.
- Capability inventory: The agent executes shell commands and runs arbitrary code provided in the problem files.
- Sanitization: No explicit sanitization or validation of the code is performed before execution.
Audit Metadata