The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the dino CLI tool exclusively for configuration management (dino config get and dino config set). The scope is restricted to the sync.timeoutMs key, minimizing the attack surface through a least-privilege approach.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a data ingestion surface by reading output from CLI commands. However, it implements strong mitigation by explicitly instructing the agent to treat all CLI output as untrusted data and prohibiting the execution of any instructions found within that data.
[DYNAMIC_CONTEXT_INJECTION]: The skill mentions the platform's !command syntax in its documentation to educate the user on manual authentication (! dino auth login). This is an informative reference to a platform feature rather than an attempt to trigger automated or hidden shell execution.
[CREDENTIALS_SAFE]: The skill includes clear directives against storing secrets in configuration files and explicitly warns against passing tokens via command-line arguments or prompts, recommending standard authentication flows instead.

dino-config