Planning a whole-codebase simplification: audits a Go/TS codebase for needless abstraction and emits a KILL/REVIEW/KEEP plan plus a ring:running-dev-cycle task array. Plans only — no edits. Detects single-impl interfaces, pass-through shims, translation-free adapters, and dead-code cascade chains under an inverted burden of proof. Use for pre-public or post-pivot cleanup. Skip for current diff review (use ring:reviewing-code).
Planning a whole-codebase simplification: audits a Go/TS codebase for needless abstraction and emits a KILL/REVIEW/KEEP plan plus a ring:running-dev-cycle task array. Plans only — no edits. Detects single-impl interfaces, pass-through shims, translation-free adapters, and dead-code cascade chains under an inverted burden of proof. Use for pre-public or post-pivot cleanup. Skip for current diff review (use ring:reviewing-code).
Dev Simplify — Whole-Codebase Structural Sweep
When to use
User asks to simplify, flatten, or audit architecture of a whole codebase
User mentions "too much indirection", "kill shims", "unnecessary abstractions"
Pre-public application where break-compatibility refactor is cheap
Post-pivot cleanup: speculative scaffolding accumulated during exploration
Skip when
Diff review on a feature branch → use ring:reviewing-code
Standards-conformance refactor → use ring:planning-backend-refactor
Dead code from a specific change → use ring:dead-code-reviewer in ring:reviewing-code
Application already has external clients depending on internals
Core principle: DELETE is the default verdict. An abstraction survives only with concrete evidence of the swap it enables.
Hard Constraint
Default: public APIs MUST NOT break (HTTP routes, SDK surface, webhooks, event contracts).
Supply hard_constraint input to override. Must be declared — never auto-inferred.
Dispatch Protocol
⛔ STOP-CHECK BEFORE DISPATCH
Before emitting any Task call, count the explorers you intend to launch in this turn.
Count MUST equal 6 (or 5 if branch has no commits ahead of main — Task 5 skipped).
If your dispatch count diverges → STOP and reconcile against the task table below.
No substitutions, no omissions.
⛔ MUST NOT trickle-dispatch
All explorers leave in the SAME TURN, before reading any explorer output.
Forbidden sequences:
Dispatch explorer 1 → read result → dispatch explorer 2
Dispatch a subset → wait → dispatch the rest
Dispatch follow-up explorers conditioned on partial output
Loop sequentially over the task list
If you find yourself about to dispatch an explorer in a turn AFTER any explorer has already returned a result → STOP. You violated parallel dispatch. Report the violation and mark the phase INCOMPLETE rather than completing the trickle.
Self-verify after dispatch
After the dispatch turn, verify all scoped Task calls (6, or 5 if Task 5 skipped) were emitted in that single turn. If fewer went out than scoped, the phase did NOT execute correctly. Mark INCOMPLETE and surface the dispatch failure — do NOT silently continue with a partial pool.
Parallel dispatch — atomic batch
Emit all scoped Task calls (the count established in the STOP-CHECK above — 6 or 5) in a SINGLE TURN, as one atomic batch.
If your runtime exposes a multi_tool_use.parallel wrapper, use it to dispatch the complete pool in one wrapped invocation. This is the canonical fan-out mechanism on OpenAI-style tool envelopes and on certain Anthropic SDK consumers — naming it explicitly activates parallel emission on runtimes where trickle-dispatch is the default behavior.
If your runtime emits parallel tool_use blocks natively (Claude Code with Claude models), multi_tool_use.parallel may not be needed — but naming it is harmless and serves as an enforcement anchor.
The STOP-CHECK, anti-trickle, and self-verify guards above remain binding regardless of which mechanism your runtime uses.
Dispatch 6 explorer agents in parallel (5 if branch has no commits ahead of main — skip Task 5):
