ratel-langsmith-analyze

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill interacts with LangSmith's well-known service endpoints (api.smith.langchain.com) to fetch observability data. This behavior is documented and aligns with the tool's core functionality.
  • [COMMAND_EXECUTION]: Employs Bash to execute Python scripts for the LangSmith SDK fallback. The scripts are used to query run statistics and trace details as a secondary access path.
  • [DATA_EXFILTRATION]: Retrieves trace inputs, outputs, and metadata from LangSmith to generate local reports in the .ratel/ directory. This data flow is restricted to the specific project scope and the local file system.
  • [PROMPT_INJECTION]: Ingests untrusted LLM transcripts from LangSmith (SKILL.md Step 4). No explicit sanitization or boundary markers are documented for the analyzed transcripts; however, the skill's capabilities (Bash, Write in SKILL.md) are used solely for generating improvement reports based on these traces.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 05:06 PM
Security Audit — agent-trust-hub — ratel-langsmith-analyze