The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions facilitate downloading and installing software from external repositories:
Fetches and executes the @mvanhorn/printing-press package from the NPM registry using npx.
Installs the digitalocean-pp-cli and digitalocean-pp-mcp tools from the github.com/mvanhorn/printing-press-library repository using go install.
[COMMAND_EXECUTION]: The skill revolves around executing a broad range of system commands via the digitalocean-pp-cli tool to manage various cloud resources, including Droplets, firewalls, and billing information.
[DATA_EXFILTRATION]: The CLI tool includes a built-in output delivery feature (--deliver webhook:<url>) that can be used to POST sensitive cloud configuration and resource data to arbitrary external URLs.
[DATA_EXFILTRATION]: A feedback mechanism is present that can be configured to transmit local data to a remote URL defined by the DIGITALOCEAN_FEEDBACK_ENDPOINT environment variable.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection due to its extensive capabilities and ingestion of external data:
Ingestion points: Processes command arguments and results retrieved from the DigitalOcean API.
Boundary markers: No specific boundary markers or delimiters are defined to isolate instructions from untrusted external data.
Capability inventory: Possesses wide-ranging cloud management permissions, local file writing capabilities, and network delivery options (webhooks).
Sanitization: No sanitization or validation logic for external content is described in the skill's instructions.

pp-digitalocean