log-triage
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted log content, creating a surface for indirect prompt injection. Ingestion points: The log_triage.py script reads text from files provided via the log_file parameter. Boundary markers: The skill does not use delimiters to isolate log content from the agent's instructions. Capability inventory: The script generates a fix plan with shell commands, and the agent has script execution enabled. Sanitization: Log data is matched via regex and included in reports without sanitization, allowing potentially malicious text to reach the agent's context.
Audit Metadata