The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions direct the agent to utilize the openship CLI for project initialization and configuration validation (e.g., openship config init, openship config validate). These are standard administrative actions for the target platform.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the user's repository, creating a potential surface for indirect prompt injection. 1. Ingestion points: SKILL.md requires inspection of package.json, lockfiles, Dockerfile, and docker-compose.yml. 2. Boundary markers: Instructions lack explicit delimiters or safety headers when processing these files. 3. Capability inventory: The agent can write configuration files and execute the platform's validation CLI. 4. Sanitization: There are no requirements for sanitizing the content of the files read from the repository.
[CREDENTIALS_UNSAFE]: Reference documentation includes example connection strings and API key placeholders. The skill provides clear guidance on using the platform's secret management features and warns against hardcoding actual credentials in the generated files.

openship-config