The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the ag2 package and associated UI extras from standard registries. It also utilizes npx to bootstrap frontend components and references a project starter repository from the vendor's GitHub organization.
[INDIRECT_PROMPT_INJECTION]: The skill implements a FastAPI endpoint that ingests user-supplied messages and context directly into the agent's processing pipeline.
Ingestion points: The message payload (type RunAgentInput) in the /chat POST endpoint defined in the run_ag_ui.py code snippet.
Boundary markers: The provided code recipes do not include specific delimiters or prompt-level instructions to ignore embedded commands within user-supplied text.
Capability inventory: The agent supports streaming text, reasoning events, and server-side tool execution as defined in the SKILL.md feature list.
Sanitization: While the code examples lack explicit sanitization, the skill's production checklist correctly advises users to validate and authorize tool inputs server-side before execution.

ag2-ag-ui