The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the dino CLI tool to perform storage operations such as listing configurations, testing connectivity, and uploading files.
Evidence: Commands like dino storage list, dino storage test, and dino storage upload are defined in SKILL.md as the canonical interface.
[INDIRECT_PROMPT_INJECTION]: The skill manages an attack surface where instructions could be embedded in external filenames or storage provider metadata, but it implements comprehensive mitigations.
Ingestion points: User-provided file paths, object keys, and structured JSON output from the dino CLI (SKILL.md).
Boundary markers: The skill contains explicit instructions to treat all CLI output and note content as untrusted and to never execute instructions found within Dinox data.
Capability inventory: The skill uses the agent's local process execution capability to run the dino binary.
Sanitization: Instructions require strict argument handling to prevent interpolation, rejection of path segments containing . or .., and the mandatory use of dry-runs before execution.

dino-storage