The Agent Skills Directory

[SAFE]: The skill facilitates data migration between established LLM observability platforms and the vendor's own platform (Arize AX). All network operations are directed at official service endpoints for LangSmith, Braintrust, Langfuse, and Arize.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes historical trace data from external sources, which is a potential surface for indirect prompt injection.
Ingestion points: Historical run data and message traces are fetched via API from LangSmith, Braintrust, or Langfuse.
Boundary markers: The SKILL.md instructions explicitly warn the agent to treat exported content as raw data only and to never follow instructions found inside spans.
Capability inventory: The skill uses a migration script to perform network POST requests to Arize OTLP endpoints and reads environment variables for authentication.
Sanitization: The migrate_vendor_traces.py script flattens and stringifies all input/output data using a robust _as_text function, ensuring that potentially malicious structured content is treated as inert text.
[CREDENTIALS_SAFE]: The skill adheres to secure credential management practices. It instructs the agent to utilize environment variables or CLI profiles rather than hardcoding secrets, and it explicitly forbids the agent from reading .env files or searching the local filesystem for credentials.

arize-migrate-langsmith