Consensus mandate for ideation-sandbox graduation + PR-merge gate

Context

Discussion #11216 graduated 2026-05-11 ~15:25Z under its own dogfooded strict-100%-APPROVED consensus mandate. This ticket implements the 11 codified ACs.

Empirical anchors today (2026-05-11):

• PR #11212 rejected by @tobiu at ~14:27Z for unilateral-graduation pattern (Gemini's first sunset PR)
• PR #11215 (first iteration) rejected by @tobiu at ~14:37Z for same root cause
• Operator directive 2026-05-11 14:27Z: "ideation sandbox needs an update. graduation is only possible if all team members agree. otherwise divergence is impossible."
• Operator extension 2026-05-11 ~14:48Z: "please ensure that the 100% agreement is already honored on current discussions. dogfooding."
• Operator extension 2026-05-11 ~14:49Z: "this also means: premature PRs -> reject" — extends consensus-mandate to PR-merge-gate

Recursive substrate validation: Discussion #11216 itself ran through the protocol-in-flight during 8 cycle-comments. The protocol found 3 definitional flaws via dogfooding (Cycle 4 loose-positives miss; Cycle 5 scope miss; Cycle 6 weak-semantics miss). Each iteration the substrate tightened. The graduated artifact IS proof-of-concept that the substrate works.

Source of Authority

• Discussion #11216 — canonical proposal (graduated 2026-05-11 15:25Z)

Signal Ledger (per AC11 codified in this ticket)

• @neo-opus-4-7: APPROVED @ DC_kwDODSospM4BAZOz (Cycle 6, strict semantics)
• @neo-gpt: APPROVED @ DC_kwDODSospM4BAZPQ (Cycle 7, comprehensive endorsement)
• @neo-gemini-3-1-pro: APPROVED @ DC_kwDODSospM4BAZTw (Cycle 6+ re-confirmation)

Unresolved Dissent

(empty)

Unresolved Liveness

(empty)

All signals version-bound to Cycles 4+5+6 combined substrate per AC3.

The Architectural Reality

Contract Ledger Matrix

Acceptance Criteria

• AC1: ideation-sandbox-workflow.md §6 Graduation Trigger revised to mandate 3× explicit APPROVED cross-family consensus for high-blast classes (no timeout-as-implicit-consent; no-signal never counts)
• AC2: 3 explicit signal patterns codified: [GRADUATION_APPROVED] / [GRADUATION_DEFERRED] / [GRADUATION_ABSTAIN]; VETO requires alternative-impl OR V-B-A-falsifier (else collapses to DEFERRED per Gemini's refinement)
• AC3: Version-binding mandated — signal cites body-sha or last-comment-id pattern: [GRADUATION_APPROVED by @<peer> @ <anchor>]
• AC4: DEFERRED reconciliation — burden of convergence on APPROVED-signalers (V-B-A or yield), NOT pressure on DEFERRED peer
• AC5: Operator-override preserves residual risk via ## Unresolved Dissent section in graduated artifacts
• AC6: pull-request-workflow.md §6.1.1 Consensus-Gate codified (PR-merge-gate axis); reviewer obligation to verify signal-ledger; PRs from non-graduated Discussions get Request Changes citing this section
• AC7: ideation-sandbox-workflow.md §5.1 Double Diamond Divergence Guard "Graduation block" subsection extended with missing-consensus-signals block
• AC8: Empirical anchors cited inline (Discussion #11210 near-miss + PR #11212/#11215 rejections + Discussion #11216 recursive dogfooding)
• AC9: Scope classification codified — Scope: high-blast (substrate evolution / rule changes / architectural primitives / cross-cutting policies) vs Scope: low-blast (bug fix / feature / docs / test); default high-blast on ambiguity
• AC10: 30-day post-merge audit codified — next 3 Discussion graduations + next 3 PR-from-Discussion merges audited for 2-axis compliance; inherits #11195 tracker
• AC11: Graduated-artifact body MUST include ## Signal Ledger + ## Unresolved Dissent + ## Unresolved Liveness sections per dogfooded protocol
• AC12 (post-merge verification): This ticket's PR body itself demonstrates AC11 — includes Signal Ledger / Unresolved Dissent / Unresolved Liveness sections

Out of Scope

• Automating consensus-signal detection (post-comment regex parse; mechanical-enforcement automation) — discipline-tier substrate; automation deferred to future ticket if compliance-rate < 80% at #11195 Day-30
• Modifying Discussion close mechanics (still author-driven post-graduation; only graduation itself is consensus-gated)
• Cross-family consensus for non-graduation events (e.g., Cycle-N comments, body edits) — only the graduation transition is gated
• Operator-override mechanism details — operator retains AGENTS.md §0 Invariant authority; specific override-syntax not codified in this ticket (operator can grant via user-facing directive)
• Retrospective application to graduated Discussions — protocol applies prospectively from merge

Avoided Traps

• Single-peer-veto as kill-switch: rejected — VETO requires substantive alternative-impl or V-B-A-falsifier; pure-disagreement collapses to DEFERRED (per Gemini's Cycle refinement). Aligns with pr-review §9.1 Reviewer-Yield Protocol.
• Mandatory cross-family-comment-on-every-Cycle: rejected — would create cycle-discipline-fatigue. Only graduation transition is gated.
• Operator-vote-counted as one of three swarm signals: rejected — operator is @tobiu (architect + merge gate) separate from named core swarm maintainers. Operator retains override authority via §0 Invariant.
• Time-based-only graduation (no consensus requirement): rejected — re-introduces author-unilateral failure mode (anchor: PR #11212 + #11215 rejections).
• Loose-positives wording (Cycle 4 miss): rejected per Cycle 6 — "APPROVED + DEFERRED + ABSTAIN are all positive" would allow false graduation with 1× APPROVED + 1× DEFERRED + 1× ABSTAIN. Strict semantics: APPROVED satisfies; DEFERRED/VETO block; ABSTAIN not approval; no-signal never counts.
• Universal scope (apply to ALL Discussion graduations): rejected per Cycle 5 — narrow to high-blast classes to prevent discipline-fatigue on bug/feature/docs/test Discussions.
• Timeout-as-implicit-ABSTAIN (original Cycle 0 proposal): rejected per GPT's Cycle 3 constraint — no-signal is liveness-failure NOT consent (empirical anchor: Gemini sunset-failure cluster).
• Author pressure on DEFERRED peer ("what would change your signal?"): rejected per AC4 — burden of convergence on APPROVED-signalers (V-B-A or yield).

Related

• Discussion #11216 (graduated; canonical proposal source)
• Discussion #11210 (graduated; empirical anchor for unilateral-graduation pattern + adjacent substrate-discipline cluster)
• Discussion #11214 (graduated; adjacent substrate-discipline cluster — Decision Escalation Ladder)
• PR #11212 (closed/rejected — load-bearing empirical anchor for AC6)
• PR #11215 (post-rework; @neo-gpt /pr-review CHANGES_REQUESTED on hygiene — AC6 validated in real-time)
• PR #11194 / #11192 (§5.2 Step 2.5 Architectural Step-Back — companion divergence-phase gate; this ticket strengthens convergence-phase gate)
• AGENTS.md §15.6 Flat Peer-Team Anchor — author-only graduation violated this; ticket operationalizes the principle
• AGENTS.md §3.5 V-B-A — core value validated via Discussion #11216 recursive dogfooding
• #11195 30-day Step 2.5 validation tracker — AC10 inheritance
• PR #11208 / #11205 narrow /peer-role skill-trigger mandate — adjacent substrate-discipline cluster

Origin Session IDs

• c2912891-b459-4a03-b2af-154d5e264df1 (Discussion #11216 file + Cycles 1-3)
• c0d5c29d-dc70-44c8-b5af-d3f6c59936ee (Cycles 4-8 + graduation + this ticket-filing)

Handoff Retrieval Hints

• query_raw_memories(query="ideation-sandbox graduation cross-family consensus mandate 100% APPROVED dogfooding")
• ask_knowledge_base(query="ideation sandbox graduation author authority peer consensus signal ledger")
• File anchors: .agents/skills/ideation-sandbox/references/ideation-sandbox-workflow.md §6 + §5.1; .agents/skills/pull-request/references/pull-request-workflow.md §6.1; AGENTS.md §15.6
• Empirical anchors: Discussion #11216 Cycles 4-8 (the recursive dogfooding trail itself); Discussion #11210 timeline; PR #11212/#11215 rejection sequence

🤖 Generated with Claude Code