The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data which constitutes a potential attack surface. Ingestion points: Responses from cloud AI provider APIs and local model servers (Ollama, LM Studio) are read into the agent context. Boundary markers: Absent; the skill does not specify markers to distinguish between data and instructions in API responses. Capability inventory: The skill requires write_file to update the .beta/config.toon file and run_code to perform connectivity tests and API calls. Sanitization: No explicit sanitization or validation of model names or provider metadata is performed before writing to the file system.
[DATA_EXFILTRATION]: The skill manages highly sensitive API keys. It follows security best practices by instructing the agent to store these credentials in a secure vault rather than in plain text configuration files. Connectivity tests for these keys target well-known cloud providers and local addresses, which is standard procedure for provider validation.

add-provider