The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as LLM traces, user-provided datasets, and evaluation rubrics, which creates a potential surface for indirect prompt injection.
Ingestion points: Data is ingested via the promptic_sdk.init() tracing mechanism, the create_observations API method, and CLI commands like promptic observations add as described in SKILL.md.
Boundary markers: The documentation does not specify explicit delimiters or instruction-ignore markers for processed data.
Capability inventory: The skill performs network operations to the official platform domain and interacts with the local file system for configuration in ~/.promptic/config.toml.
Sanitization: No specific sanitization or filtering logic for ingested data is detailed in the provided instructions.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the promptic-sdk package and its various provider-specific extras (such as openai, anthropic, and langchain) from the official Python package registry.
[COMMAND_EXECUTION]: The skill utilizes the promptic CLI tool for managing authentication, workspace configuration, and triggering agent evaluations.

promptic