The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to research application deployment facts by fetching content from external sources such as official documentation, LinuxServer.io, or container registry image pages.
Ingestion points: Technical details (images, ports, environment variables) are retrieved from third-party websites via web fetch in the 'Flesh out the idea' step.
Boundary markers: The skill does not provide specific instructions or delimiters to isolate fetched external content or prevent the agent from following instructions embedded within those external pages.
Capability inventory: The agent has the capability to write to the local filesystem (compose.yaml, .env), execute validation commands (docker compose config), and perform version control operations (git add, git commit) as detailed in the 'Write the files', 'Validate', and 'Commit and hand off' sections.
Sanitization: There are no requirements for the agent to sanitize or validate the technical specifications retrieved from the web before using them to generate executable Compose files or scripts.
[COMMAND_EXECUTION]: The skill uses shell commands to perform configuration validation and repository management.
Evidence: Instructions include running docker compose --project-directory apps/<app-name> config -q for validation and git add / git commit for repository updates.

arcane-compose