The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is centered around the use of the difyctl CLI to discover and execute operations on a Dify server. It instructs the agent on how to map server operations to terminal commands and handle JSON-based results.
[CREDENTIALS_UNSAFE]: The documentation describes the use of DIFY_TOKEN for authentication and notes that difyctl login can store tokens in local files when the --no-keyring flag is used. It also mentions the --insecure flag, which allows the agent to skip TLS verification when communicating with a server.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process hints and input_schema returned from the remote Dify server to decide on subsequent steps, such as form submissions or following a reply operation. This creates a potential surface where a compromised or malicious server could provide instructions that influence the agent's logic.
Ingestion points: JSON responses from the Dify server parsed via the difyctl CLI (SKILL.md).
Boundary markers: The skill lacks explicit instructions to sanitize or validate the intent of the 'hints' provided by the server before the agent acts on them.
Capability inventory: Access to the difyctl CLI which can perform server-side runs, file uploads, workspace management, and authentication configuration.
Sanitization: While the skill mentions using the server's input_schema for structure, it does not define sanitization for the natural language content within the hints or error messages.
[EXTERNAL_DOWNLOADS]: The skill references the difyctl install skills command, which is used to install additional agent capabilities from a provided path, facilitating the expansion of the agent's toolset from external sources.

difyctl