The Agent Skills Directory

[CREDENTIALS_UNSAFE] (LOW): The skill documents the storage of API tokens in ~/.config/omnivore-api/config.yaml. While standard for CLI tools, this creates a sensitive file on the local filesystem.
[DATA_EXFILTRATION] (LOW): The skill interacts with the external domain api-prod.omnivore.app to fetch and save user data. This is consistent with its primary purpose but involves transmitting information to a third-party service.
[PROMPT_INJECTION] (LOW): High risk of Indirect Prompt Injection (Category 8).
Ingestion points: omnivore get-articles fetches external content from the web via the Omnivore service.
Boundary markers: None mentioned in the documentation; article content is processed directly.
Capability inventory: The skill allows command execution and file modification (omnivore_api/cli.py).
Sanitization: No evidence of sanitization for fetched article content before it reaches the agent context.
[COMMAND_EXECUTION] (SAFE): The skill executes the omnivore CLI tool and provides instructions for the agent to extend its functionality by writing Python code using the typer library.

omnivore-cli