What problem does it solve? Maintaining a LangGraph-based scrum assistant requires consistent patterns across agent nodes, prompts, tools, LLM providers, and state schemas; this Skill encodes those conventions so changes to src/yeaboi/agent/, prompts/, tools/, and sessions.py stay compatible with saved sessions and existing tests. ## Core Features & Use Cases - Node and prompt conventions: Enforces the parse→fallback→format node pattern, auth/billing error re-raising, human-in-the-loop review routing, and ARC-framework prompt factories. - LLM provider and tool registration: Documents lazy-imported provider setup (Anthropic, OpenAI, Google, Bedrock, Ollama, OpenAI-wire vendors) and the single get_tools() registration pattern. - State schema and session persistence: Covers ScrumState TypedDict rules, frozen-dataclass backward compatibility, schema versioning (v12), and dict_to* deserialization for --resume. - Use Case: When adding a new field to a frozen dataclass like Feature, follow the Skill's rule to give it a default value so old saved sessions still deserialize, then add tests in test_state.py. ## Quick Start Ask the agent to add a new optional field to the ScrumState in agent/state.py following the project conventions, including a default value and a corresponding test.
