effort-worthiness — rigor must buy goal progress

Census id: F13 · Family: F — rules corpus + generated-rule tooling + shipped audit scripts · Provenance: framework pin b069c59328aa8e08671d3a19fd25efa618dfc1c5 (staging FETCH_HEAD, read 2026-09-11; re-anchored from census pin a1337cb301). Every file:line below is at that pin. Raw reference drafted by stage E4 — factual, unpolished by design.

What it bans

Effort theatre: rigor spent to prove effort rather than move the goal. .claude/rules/effort-worthiness.md:1 «# Effort worthiness — rigor must buy goal progress»; the default and burden: :27 «Default is practice-first: build the reversible thing and verify it live. The burden of» / «proof sits on MORE rigor — whoever demands a probe, an experiment, or an extra round must» / «state (a) what breaks if we skip it and (b) what learning-in-practice costs instead» / «(proportionality rule).» Class header: :8 «> Class: C — prose statute; the mechanical skeleton lands on its own channels (§2 L3).» The KPI line: :47 «Zero-finding reviews and «not worth it» verdicts are legitimate outcomes — the KPI is» / «goal-shift, never findings-produced or rounds-run.»

Never (fires)

The named anti-patterns: .claude/rules/effort-worthiness.md:74 «- #type1-process-on-type2 — heavy research-grade process on a reversible surface; the named failure mode of the Type-1/Type-2 distinction (§4). Counter: L0 label + L1 test 4.»; :76 «- #findings-as-KPI — a review round or follow-up PR spawned to prove effort, carrying no failure scenario. Counter: severity contract»; :78 «- #rigor-by-paraphrase — re-litigating a recorded verdict without new evidence. Counter: L2 journal + copy-or-pointer transfer.»

Always (clean)

The four-test card at every effort fork, quoted: .claude/rules/effort-worthiness.md:34 «Before spending effort, run the card — four tests, judged honestly, trace recorded:» then :36-39 («1. Does this effort move us toward the goal? 2. Is it theatre (form satisfied, substance absent)? 3. Is it immaterial — changes nothing a consumer or decision would notice? 4. Material but cheaper to verify in practice after building?»). Materiality scaling: :41 «Materiality scales with the highest layer touched (idea → design → architecture → plan → implementation; conflicts resolve upward)».

Enforcement channels

• Rule-index row: .claude/rules/00-rule-index.md:21 «| effort-worthiness.md | C | any effort/rigor fork: probe demand, extra round, budget breach. | skill-embed(4) |»
• Skill-embed, four hosts: .claude/rules/effort-worthiness.md:3 «» through :6 (night-mode, arch, reviewer-discipline).
• L3 mechanical skeleton, named channels (:59-64): «(a) Failure-scenario: required on round-triggering findings — deterministic arm in [pr-body-fidelity.ts] …; (b) ask-file schema validity + (c) answered⇒journal-entry cross-check — pre-push section …; (d) L0 label presence — kickoff principle test (landing item).»
• Not backend-rendered — no FF diagnostic applies (FF7001/FF7002, packages/core/diagnostics/registry.ts:321/:329); honest status: four skill-embeds + deterministic pr-body-fidelity arm; promotion criteria per layer at :106-109.