The Agent Skills Directory

[SAFE]: The skill provides instructions and code samples for using the 'langwatch' SDK and CLI to perform evaluations. These are legitimate resources owned by the skill's author.
[SAFE]: No obfuscation, data exfiltration, or unauthorized command execution was detected. All shell commands are documented as part of the tool's intended help and documentation retrieval functionality.
[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing and evaluating untrusted inputs such as images, audio, and PDFs.
Ingestion points: Data enters the context via image_dataset loops and file analysis commands in SKILL.md.
Boundary markers: None explicitly defined in the provided snippets.
Capability inventory: The skill demonstrates usage of the langwatch Python SDK for experiment tracking and evaluation.
Sanitization: Not present in the evaluation logic provided.

evaluate-multimodal