The Agent Skills Directory

[SAFE]: The skill does not exhibit any malicious patterns. It provides a structured framework for managing complex agent workflows with explicit boundaries and state management.
[EXTERNAL_DOWNLOADS]: The skill implements a version check that performs an anonymous GET request to the official GitHub repository metadata endpoint. This is a read-only check used to notify the user of available updates and does not involve downloading or executing remote code.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured data generated by language models (e.g., retrieval plans and judgment criteria). It mitigates potential injection risks by using dedicated validation functions that enforce schema constraints and unique identifiers before the data influences subsequent execution stages.
[COMMAND_EXECUTION]: The provided examples use standard environment variable access for local configuration and utilize the asyncio library for managing concurrent tasks without spawning unauthorized external processes or elevated privileges.

agently-triggerflow