The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of AI agents that process user-supplied messages and tool outputs, creating a standard surface for indirect prompt injection.
Ingestion points: The RunAgentInput and ClientState structures in ts/agui-protocol.md and ts/adapter-langgraph.md define how external data, including messages and tools, enter the agent's context.
Capability inventory: The skill enables agents to perform network operations via SSE streaming and execute tools both on the server and the client side (ts/agui-protocol.md).
Boundary markers: The documentation does not explicitly detail the use of delimiters for untrusted input; however, it leverages established frameworks like LangGraph and LangChain which have their own integration patterns.
Sanitization: Not explicitly addressed in the snippets, implying reliance on the developer's implementation or the underlying framework's safety features.
[EXTERNAL_DOWNLOADS]: The skill references several official packages from the @cloudbase and @ag-ui scopes (e.g., @cloudbase/agent-server, @cloudbase/agent-adapter-langgraph, @ag-ui/client). These are vendor-provided resources from the author 'tencentcloudbase' and are considered safe.
[COMMAND_EXECUTION]: The documentation includes standard deployment commands such as setting executable permissions (chmod +x scf_bootstrap) and using the manageAgent MCP tool for cloud deployment. These operations are restricted to the intended deployment workflow and show no signs of malicious intent.

cloudbase-agent