split-initiative

Read first

• ../references/initiative-packet-contract.md
• ../references/initiative-workflow-contract.md
• ../references/final-state-authoring-policy.md
• initiative BRIEF.md, UBIQUITOUS_LANGUAGE.md, DECISIONS.md, INVARIANTS.md, REPO_MAP.md, CONTRACTS.md, OPEN_QUESTIONS.md
• optional SPLIT_OPTIONS.md
• ../../references/communication-mode.md

Goal

Write the first semantic TASK_GRAPH.md so each child is independently understandable, has explicit dependencies, and is small enough for a normal Atelier child loop.

Heuristics to internalize

• prefer vertical slices over horizontal phase buckets
• make each child understandable without giant hidden context
• state explicit scope, out-of-scope, validation expectations, and dependencies
• sequence by uncertainty and risk reduction
• keep one main change seam per child when possible
• prefer one clear owner repo per child; treat truly multi-repo children as a smell unless the boundary is genuinely inseparable

Write into TASK_GRAPH.md

For each child:

• slug / title
• owner or owning repo / boundary
• Packet Root: repo root where the child packet will live
• dependency edges
• scope summary
• contract surface summary
• validation expectations
• optional blocker note
• Child Status: ready when dependencies are already satisfied, otherwise Child Status: pending

Status

• enter initiative Status: split_drafting if no split-drafting phase already exists
• exit with initiative Status: split_review

Writing rule

Write the task graph as the canonical execution decomposition. Do not leave superseded split vocabulary or transitional child-state names in the graph.

Do not

• create giant catch-all children
• use the parent packet as a delivery plan
• reference outside Atelier skill files for split heuristics

Communication

Honor active caveman mode for user-facing replies per ../../references/communication-mode.md. Keep durable artifacts normal unless the human asks otherwise. Drop caveman for safety/clarity when needed, then resume.