The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the langsmith Python package and retrieves prompt templates from the LangSmith Hub (e.g., langchain-ai/rag-prompt). These resources originate from well-known and established service providers.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose involves processing LLM inputs and outputs, which creates a surface for indirect prompt injection. Ingestion points: Tracing decorators in SKILL.md and references/advanced-usage.md capture external LLM inputs/outputs. Boundary markers: Documentation includes guidance on input filters to manage untrusted data. Capability inventory: Transmission of captured data to the LangSmith observability platform. Sanitization: Code examples demonstrate masking sensitive fields like passwords using process_inputs before data transmission.
[SAFE]: Sensitive configuration, such as API keys, is handled through standard environment variable placeholders. Advanced features like filesystem attachments and SSL verification overrides are documented with appropriate security context and warnings in the troubleshooting and advanced usage guides.

langsmith-observability