drift-detection
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection due to its role in processing external LLM trace data. * Evidence: 1. Ingestion points: Data is ingested through langfuse.fetch_scores and langfuse.get_traces in SKILL.md and integration references. 2. Boundary markers: No delimiters or instructions to ignore embedded prompts are implemented. 3. Capability inventory: The skill is limited to statistical calculations; it contains no subprocess calls, file-write operations, or network exfiltration logic. 4. Sanitization: No sanitization or filtering of text data is performed prior to analysis.
- [EXTERNAL_DOWNLOADS] (LOW): The skill requires third-party packages for data fetching and monitoring. * Evidence: Use of langfuse, evidently, pandas, and numpy is documented for statistical processing and platform integration.
Audit Metadata