The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages sensitive credentials, including API keys, passwords, and environment secrets, which are stored in local profiles and transmitted to the platform API through commands like duvo login, duvo secrets, and duvo credentials.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the @duvoai/cli package from npm and includes a self-update command (duvo update) that fetches new versions of the binary from the vendor's distribution channels.
[COMMAND_EXECUTION]: The skill makes extensive use of the duvo CLI to manage platform resources and includes a low-level "api escape hatch" (duvo api) that allows for arbitrary HTTP requests to the platform's public endpoints, providing high capability to the agent.
[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to process data from external sources and perform impactful actions, creating an attack surface where malicious data could influence agent behavior.
Ingestion points: Processes case data via cases list, run messages via runs messages, and file contents via files content or sandboxes files.
Boundary markers: The instructions do not specify the use of delimiters or "ignore embedded instructions" warnings when processing data from the platform.
Capability inventory: Includes destructive operations such as agents delete, runs stop, secrets delete, and arbitrary API access via the duvo api command.
Sanitization: No explicit sanitization, validation, or escaping of external content is documented before the data is interpolated into commands or agent context.

duvo-cli