The Agent Skills Directory

[PROMPT_INJECTION]: The skill explicitly instructs the AI agent to skip user confirmation for the deployComposeApp tool, directing it to call the tool "in the SAME turn" and stating "don't ask the user to confirm first." This attempts to override standard agent safety protocols requiring human-in-the-loop oversight for significant actions.\n- [EXTERNAL_DOWNLOADS]: The skill retrieves template files and source code from the author's GitHub repository using npx degit and git clone. Evidence found in Step 0b: npx -y degit goldsky-io/documentation-examples/compose/VRF.\n- [COMMAND_EXECUTION]: The skill requires the agent to execute various shell-based commands for environment management, including the use of goldsky CLI, jq, git, and npx. It includes commands for directory cleanup (rm -rf .compose/) and secure file deletion (shred -u /tmp/k.json) to manage local cache and protect temporary key material.

compose-vrf