The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates a temporary Python script (.py) outside the repository containing logic to run the opik_optimizer. This script is then executed by the agent to perform the optimization run.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of datasets (opik.Dataset) and prompts from the library or codebase. These inputs are used as training/validation data for a reasoning model during optimization, creating an attack surface where malicious data could influence the agent's behavior or the optimization outcome.
Ingestion points: The skill reads datasets via client.get_dataset and prompts via client.get_chat_prompt or local file Read operations.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for the ingested data.
Capability inventory: The skill utilizes Bash for script execution, Write for runner creation, and Read/Grep for data retrieval.
Sanitization: The instructions do not specify any validation or sanitization steps for the ingested prompt text or dataset items.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the opik-optimizer package using uv add or pip install. This is a vendor-owned package associated with Comet ML.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute the generated Python runner and to install necessary dependencies.

opik-optimize