The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone a repository from a third-party GitHub account (https://github.com/tashfeenahmed/freellmapi.git) and install numerous dependencies using npm install.
[REMOTE_CODE_EXECUTION]: The installation process involves executing downloaded code from an unverified source to run the proxy server and dashboard.
[COMMAND_EXECUTION]: The skill uses shell commands for setup, environment configuration, and service management, including npm run build, node server/dist/index.js, and pm2 for persistence.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The proxy ingests natural language and structured data from approximately 14 external LLM providers (e.g., Google, Groq, Mistral) via their respective APIs.
Boundary markers: The instructions do not specify any delimiter or boundary enforcement to prevent the agent or calling application from following malicious instructions potentially embedded in external provider responses.
Capability inventory: The skill performs network operations (fetch) to contact LLM providers and has local file system access to manage an encrypted SQLite database and environment files.
Sanitization: The skill documentation mentions logic to translate tool call formats for compatibility (e.g., for Google Gemini) but lacks details on sanitization of response content to prevent injection.

freellmapi-proxy