alanblythe/agent-platform-skills

This skill should be used when debugging a deployment on Gemini Enterprise Agent Platform: 401 or 403 between agents, "permission denied" on an Agent Runtime agent, "CREDENTIALS_MISSING", "Identity Pool does not exist", an IAM grant that changes nothing, an agent answering confidently with invented data, a deploy failing on an undocumented permission, a 404 for a model that exists, or a "global" vs regional location mix-up. Covers identifying the real caller, why 401 and 403 need opposite responses, Agent Identity principal forms, which region belongs where, and the conditions under which a broken deployment reports success. Use agent-platform-architecture instead for choosing a design.

This skill should be used when designing how agents fit together on Gemini Enterprise Agent Platform: "should this be one agent or several", "how do I expose an agent in Gemini Enterprise", "do I need Agent Gateway", "should I enable Agent Identity", "can these agents call each other", "how do agents authenticate to each other", "how does an agent call a Cloud Run service", "why does A2UI not render", or when choosing between Agent Runtime and Cloud Run. Gives the platform constraints, the transport compatibility matrix for agent-to-agent calls, and the architectures they permit. Use agent-platform-a2a for debugging a built deployment.

This skill should be used when a multi-agent build on Gemini Enterprise Agent Platform runs without erroring but is quietly wrong: "the agent invented data the tool returned correctly", "it works with curl but not in Gemini Enterprise", "my sub-agent returned 200 and the caller still made something up", "a tool argument is fabricated", "the orchestrator dropped half the list", "/run_sse returns 404", "how do I read a sub-agent's output", "how do I stop the model inventing tool arguments", "my eval score did not change", "eval grade scored the wrong run", "the metric scores 0 on a working agent", "my multi-turn eval case tests something that never happened", "the A2UI card renders but the button does nothing", "which A2UI components does Gemini Enterprise actually render", "one component of my surface is missing", "the model keeps re-reading its own surface", "how do I build an A2UI tree", or "a parallel stage produced no output and no error". Covers what actually crosses an agent-to-agent boundary, guards that a model cannot game, building an A2UI surface that survives the round trip, and the verification traps that report a score unrelated to your agent. Use agent-platform-architecture to choose a host or transport — including whether A2UI can render there at all — agent-platform-a2a for calls failing with a status code, and google-agents-cli-eval for eval methodology, dataset schema and metric choice.

This skill should be used when deciding where data lives on Gemini Enterprise Agent Platform, across Agent Platform Sessions and Agent Platform Memory Bank: "how do I persist X", "does this survive a restart", "should I use Sessions or Memory Bank", "can agents share state across users", "how do I share memories across users", "why is my state empty in a new conversation", "what does app: or user: prefixed state do", "how do I list every memory in a scope", "why did my tool write to the wrong scope", "why is the console's Sessions page blank", "why is my agent's whole run one event", "which location does VertexAiSessionService take", or when a value written by a tool is missing on the next turn or the next instance. Names each managed service, gives its scope and lifetime, and maps it to the ADK client that binds to it. Use agent-platform-architecture for choosing a host or identity mode, agent-platform-a2a for auth failures.