roblox-debug
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the reading of external code and subsequent file writing, which could allow malicious instructions within the scripts to affect agent execution.
- Ingestion points: Untrusted source code files read in Step 2.
- Boundary markers: None present to distinguish between data and instructions.
- Capability inventory: File system read/write access via project synchronization and MCP tools.
- Sanitization: No filtering or escaping of ingested code content is implemented.
Audit Metadata