The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the local openship CLI and git binaries to manage projects, services, and deployments. Commands are executed as discrete argument sequences without a shell to prevent injection attacks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files, build logs, and environment variables. It implements a multi-layered defense strategy including mandatory preflight scans for secrets, pattern-based leak detection in configurations (e.g., detecting printenv or set -x), and automated output sanitization that redacts tokens and private keys before they enter the agent's context.
[SAFE]: The skill demonstrates high security maturity by explicitly forbidding access to local credential files (e.g., ~/.openship/config.json), using fail-closed logic for sensitive operations, and providing dedicated scripts to ensure that secrets are never echoed in commands or logs.

openship