The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a CLI tool cyeam to interact with the Trello API. All interactions (creating cards, moving lists, downloading attachments) are mediated through this binary rather than direct HTTP requests, which is a standard method for extending agent capabilities.
[CREDENTIALS_SAFE]: The skill documentation explicitly instructs the agent to store Trello tokens in the system keychain or a restricted file (~/.cyeam/trello.json with 0600 permissions). It also contains a strict rule: "Never output credentials."
[DATA_EXFILTRATION]: While the skill downloads Trello attachments, it specifically warns against sending protected Trello URLs to external platforms like Lark/Feishu, requiring the agent to use local file paths (saved_to) for authorized transfers. This ensures that sensitive file access remains within the authenticated session context.
[PROMPT_INJECTION]: No malicious prompt injection patterns or attempts to bypass safety filters were detected in the skill instructions.

trello