Decision 003: Bootstrap a v0 skeleton first, defer runtime automation

• Status: accepted
• Date: 2026-06-27
• Deciders: @MadaraUchiha-314 (via issue #1), the-loop (self)
• Work item: issue-1

Context

Issue #1 is a broad vision spanning ticketing, tooling, multi-party collaboration, the execution loop, docs/decisions/learnings, self-improvement, distribution, webhooks and remote execution. It cannot be delivered in a single pass, and the repo starts empty. The issue's final directive: "the-loop's first task is to create itself."

Decision

Deliver a coherent v0 foundation that establishes the structure and contracts of the-loop, and defer runtime automation to follow-up work items:

• In v0: plugin + marketplace manifests; config schema + default config; manifest; epic/story/bug + plan/log/decision/learning templates; init / work-on / upgrade-the-loop commands; the the-loop skill; a SessionStart hook; docs (architecture, decisions), learnings, and the self-referential delivery plan + execution log for issue #1.
• Deferred (future issues): webhook triggers (PR comments, Actions), remote-workspace execution, DAG orchestration across work items, concrete language-specific tooling integrations (uv/bun/nx/pytest/vitest/playwright/oxlint/ ruff/pyright), messaging integrations, and Cursor packaging.

Consequences

• The user gets a reviewable, installable skeleton quickly, embodying every rule from the issue as documented structure.
• Each deferred capability becomes its own work item delivered through the-loop itself.

Alternatives considered

• Attempt the whole vision at once — rejected: not deliverable or reviewable.
• Build only docs — rejected: leaves nothing installable/usable.