The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to install the @acedatacloud/ai-chat extension via the clawhub package manager. This is a vendor-specific resource intended to provide the documented chat capabilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process external and potentially untrusted data—such as audio files, news summaries, and implementation plans—creating a surface for indirect prompt injection.\n
Ingestion points: Untrusted data enters the agent context through the message, question, and references parameters within the /aichat2/conversations and Gemini native API endpoints documented in SKILL.md.\n
Boundary markers: The provided documentation and examples do not define specific delimiters or instructions to the model to ignore embedded commands within user-supplied content.\n
Capability inventory: The skill utilizes network operations (via curl and the openai Python SDK) to communicate with api.acedata.cloud and accesses local configuration files in ~/.openclaw/.env.\n
Sanitization: No explicit content sanitization or validation logic is mentioned for the data forwarded to the LLM endpoints.

ai-chat