What problem does it solve?
It eliminates inconsistent and fragile integrations between Child Agents and the AgentHub ADK Runtime by defining a single, contract-first development contract for configuration, streaming, artifact output, tool permissions, error handling, and observability.
Core Features & Use Cases
- Unified runtime-to-protocol mapping: Ensures handlers emit A2A-compatible lifecycle events via ADK Context rather than directly constructing protocol payloads.
- Explicit capability and safety contracts: Forces Child Agents to declare skills, input/output modes, and tool permissions through config.yaml and AgentCard, with a default-minimal permission posture.
- Consistent streaming and artifact production: Standardizes ctx.StreamText for text streaming and ctx.AddArtifact for ArtifactDraft outputs while preventing leakage of platform-specific fields into drafts.
- LLMClient lifecycle and security guardrails: Requires LLMClient creation at startup, HTTP timeouts, secret hygiene, error de-sensitization, and stable error codes/log fields.
- Review and contract testing checklist: Provides a concrete review rubric to catch boundary violations (no SSE/HTTP/AG-UI event generation in handlers, no gateway planning/registry access, artifact schema compliance, observability coverage).
Quick Start
Use this skill to implement a new Child Agent runtime integration that loads config.yaml, validates/generates AgentCard, exposes /health and an A2A server, and routes tasks through a handler that only uses ctx.Context/StreamText/AddArtifact/Fail while adhering to the review checklist.
