The Agent Skills Directory

[SAFE]: The skill performs network requests to openrouter.ai, which is a well-known and established service in the AI ecosystem. These operations are required for model metadata discovery and chat completions, representing the primary intended functionality of the skill.
[SAFE]: Data persistence is managed via a local cache file at ~/.opencode/openrouter-model-cache.json. The implementation employs secure file permissions (0o600), ensuring that cached model data is only accessible by the current user, which adheres to security best practices.
[SAFE]: Secret management for OPENROUTER_API_KEY relies on standard environment variable practices. The skill provides clear instructions for users to set their own keys and does not contain hardcoded credentials or mechanisms to exfiltrate these secrets to unauthorized domains.
[SAFE]: The skill includes an automated initialization hook (onAgentReady) that runs its own bootstrap command. This is a standard lifecycle management pattern used to ensure the model registry is populated when the agent starts, with no evidence of malicious command injection.

openrouter-free