oh-my-customcodex 0.5.7 → 0.5.9

package/templates/guides/agentmemory-migration/phase-1-coexist.md

@@ -1,261 +1,27 @@
-# AgentMemory Migration — Phase 1: COEXIST (Week 1-2)
+# External Memory Migration — Retired COEXIST Plan
 
-> **상태**: 활성 — 2026-05-18 시작 (#1169)
-> **다음 단계**: Phase 2 SWITCH (measure 결과 후 GO/NO-GO 결정)
-> **이전 단계**: [measure-step-zero.md](./measure-step-zero.md)
+> **Status**: Retired by #1426. COEXIST mode is no longer supported for this project.
+> **Current policy**: Native auto-memory first; optional `omx-memory`/AgentMemory-compatible searchable backend only when already configured.
 
----
+## Why This Plan Is Retired
 
-## 1. 개요
+The old coexistence plan introduced dual-backend complexity and session-end save ambiguity. The current project decision removes that complexity: deprecated Chroma-based memory tooling is not installed, invoked, measured, or treated as a fallback.
 
-Phase 1 (COEXIST)는 claude-mem과 AgentMemory를 **동시에 운영**하는 단계입니다.
-이 단계에서는 어떤 destructive 변경도 발생하지 않습니다:
+## Active Policy
 
-- claude-mem 기존 Chroma 데이터 유지
-- 12 plugin skill 폐기 없음
-- 어댑터 코드 활성화 없음 (STUB 상태 유지)
-- 자산 처리표 적용 없음
+| Need | Supported path |
+| --- | --- |
+| Agent-local durable facts | Native `MEMORY.md` via agent frontmatter `memory:` |
+| Cross-session search | `omx-memory` or AgentMemory-compatible MCP exposing `memory_search`/`memory_read` |
+| Session summary save | `memory_add` or `observation_add` when an approved backend is configured |
+| No approved backend | Skip searchable save and continue; memory failures do not block the main task |
 
-목표는 1주간 실제 사용 데이터를 수집하고, AgentMemory 설치와 공존 운영에 익숙해지는 것입니다.
+## Session-End Checklist
 
-| 항목 | Phase 1 (COEXIST) | Phase 2 (SWITCH) |
-|------|-------------------|------------------|
-| claude-mem | 활성 (기존 운영) | 비활성화 예정 |
-| AgentMemory | 설치 후 병렬 활성 | 단독 운영 |
-| 데이터 마이그레이션 | 없음 | 선택적 (측정 후 결정) |
-| 어댑터 코드 | STUB 유지 | 활성화 |
-| 자산 처리표 | 적용 보류 | 사용자 검토 후 적용 |
+1. `sys-memory-keeper` updates native memory when a session-end signal exists.
+2. Orchestrator attempts approved searchable memory save only if the tool is available.
+3. Deprecated plugin-specific wrappers and Chroma tool calls remain unused.
 
----
+## Follow-up
 
-## 2. AgentMemory 설치
-
-> **R001 주의**: auto-install 금지. 아래 명령어를 사용자 로컬에서 수동으로 실행합니다.
-
-### 2-1. 패키지 설치
-
-```bash
-# 옵션 A: pip
-pip install agentmemory
-
-# 옵션 B: pipx (격리 환경 권장)
-pipx install agentmemory
-
-# 옵션 C: uvx (uv 사용 시)
-uvx agentmemory
-```
-
-### 2-2. MCP 서버 등록
-
-```bash
-# Claude Code CLI로 등록
-claude mcp add agentmemory -- agentmemory mcp
-
-# 또는 .mcp.json 수동 편집 (R001 auto-install 금지)
-# {
-#   "mcpServers": {
-#     "agentmemory": {
-#       "command": "agentmemory",
-#       "args": ["mcp"]
-#     }
-#   }
-# }
-```
-
-### 2-3. 등록 확인
-
-```bash
-claude mcp list
-# 출력에 agentmemory 서버가 표시되어야 합니다
-```
-
----
-
-## 3. COEXIST 정책
-
-Phase 1 기간 동안 두 backend는 다음 정책에 따라 공존합니다.
-
-### 3-1. 기본 운영 방식
-
-| 작업 | 우선 backend | 비고 |
-|------|-------------|------|
-| 기존 메모리 조회 | claude-mem | 기존 Chroma 데이터 유지 |
-| 신규 메모리 저장 | 둘 다 | COEXIST 기간 중 병렬 저장 권장 |
-| 세션 종료 시 저장 | 둘 다 | R011 Session-End Self-Check 참조 |
-
-### 3-2. 쿼리 병합 (memory-aggregator)
-
-두 backend에서 결과가 반환될 경우 `memory-aggregator` 스킬이 결과를 병합합니다:
-
-```
-claude-mem.search(query)    →  결과 A
-AgentMemory.search(query)   →  결과 B
-memory-aggregator           →  A + B 중복 제거 후 반환
-```
-
-`memory-aggregator` 스킬이 없는 경우, 두 결과를 순서대로 제시하고 사용자가 선택합니다.
-
-### 3-3. 충돌 방지 원칙
-
-- claude-mem 기존 데이터를 AgentMemory로 자동 복사하지 않음
-- 두 backend에 동일 키로 저장 시 별개 항목으로 취급
-- 강제 동기화 없음 — Phase 2에서 마이그레이션 여부 결정
-
----
-
-## 4. Dual-Backend 충돌 Advisory
-
-`.mcp.json`에 두 서버가 동시 등록된 상태를 감지하면 다음 advisory를 출력합니다.
-
-### 4-1. 감지 조건
-
-```
-.mcp.json 내 서버 목록:
-  - "claude-mem" (또는 관련 tool prefix)
-  - "agentmemory"
-  → 두 개 동시 존재 시 Phase 1 COEXIST 상태로 간주
-```
-
-### 4-2. Advisory 메시지
-
-```
-[Advisory] Dual memory backend detected (Phase 1 COEXIST)
-  - claude-mem: active (Chroma)
-  - agentmemory: active (SQLite)
-  현재 Phase 1 COEXIST 정책 적용 중 — 두 backend 병렬 운영
-  Phase 2 SWITCH 진입 전까지 두 backend 유지
-  가이드: guides/agentmemory-migration/phase-1-coexist.md
-```
-
-이 advisory는 경고가 아닙니다. Phase 1에서는 정상 상태입니다.
-
-### 4-3. 강제 선택 (사용자 명시 시)
-
-사용자가 명시적으로 "agentmemory만 사용" 또는 "claude-mem만 사용"을 요청하는 경우,
-해당 요청을 따르되 자산 처리표 적용 및 어댑터 활성화는 Phase 2 GO 결정 후 진행합니다.
-
----
-
-## 5. 사용자 행동 계획
-
-### 5-1. 1주 measure 루틴 (자동 트리거: 2026-05-25)
-
-`scripts/measure-claude-mem-usage.sh` 스크립트를 1주간 실행합니다:
-
-```bash
-# 수동 실행 (필요 시)
-bash scripts/measure-claude-mem-usage.sh
-
-# 결과는 .claude/outputs/sessions/ 에 저장됩니다
-```
-
-측정 항목:
-- 일별 호출 횟수 (claude-mem vs AgentMemory)
-- 응답 지연 (p50, p95)
-- 저장 성공/실패율
-- 메모리 용량 (Chroma vs SQLite)
-
-### 5-2. GO/NO-GO 결정 기준 (2026-05-25 예정)
-
-| 지표 | GO 조건 | NO-GO 조건 |
-|------|---------|------------|
-| 응답 지연 | AgentMemory ≤ claude-mem × 1.2 | AgentMemory > claude-mem × 2.0 |
-| 저장 성공률 | ≥ 99% | < 95% |
-| 운영 안정성 | 1주 무장애 | 크래시 또는 데이터 손실 |
-| 호환성 | 기존 메모리 포맷 읽기 가능 | 포맷 비호환으로 검색 실패 |
-
-### 5-3. 지금 할 수 있는 것
-
-- AgentMemory 설치 및 MCP 등록 (위 섹션 2 참조)
-- 신규 세션에서 AgentMemory로도 저장해보기 (실험적)
-- measure 스크립트 결과 관찰
-- 자산 처리표 (`guides/agentmemory-migration/asset-disposition.md`) 검토 준비
-
----
-
-## 6. Phase 2 진입 조건
-
-다음 조건이 모두 충족된 경우에만 Phase 2 (SWITCH)로 진행합니다.
-
-### 필수 조건
-
-- [ ] 1주 measure 결과 GO 판정 (5-2 기준 충족)
-- [ ] 자산 처리표 사용자 검토 완료 (12 plugin skill 처리 방향 결정)
-- [ ] 30분 롤백 절차 검증 완료
-
-### 롤백 절차 검증 (Phase 2 진입 전 필수)
-
-```bash
-# 1. Chroma 백업 생성
-cp -r ~/.local/share/claude-mem/chroma ~/.local/share/claude-mem/chroma.bak.$(date +%Y%m%d)
-
-# 2. 백업에서 복원 가능 여부 확인
-# (실제 복원 수행하지 않고 절차만 검증)
-ls -la ~/.local/share/claude-mem/chroma.bak.*
-
-# 3. AgentMemory SQLite 백업
-cp ~/.local/share/agentmemory/memories.db ~/.local/share/agentmemory/memories.db.bak.$(date +%Y%m%d)
-```
-
-롤백 소요 시간이 30분 이내임을 확인한 후 Phase 2 진행.
-
----
-
-## 7. 현 단계 한계 및 금지 사항
-
-### 7-1. STUB 상태 유지 (변경 금지)
-
-```
-packages/eval-core/src/adapters/agentmemory.ts  ← STUB, 활성화 금지
-packages/eval-core/src/db/schema.ts              ← 변경 금지
-```
-
-이 파일들은 Phase 2 GO 결정 후 별도 사이클에서 수정합니다.
-
-### 7-2. 자산 처리표 적용 보류
-
-12 plugin skill의 폐기/유지 결정은 measure 결과를 확인한 후 진행합니다:
-
-```
-.claude/skills/memory-*/  ← 폐기 여부 보류
-.claude/skills/claude-mem-*/  ← 폐기 여부 보류
-```
-
-### 7-3. 데이터 마이그레이션 금지
-
-Phase 1에서는 claude-mem Chroma 데이터를 AgentMemory SQLite로 이전하지 않습니다.
-이전은 Phase 2에서 사용자가 명시적으로 승인한 경우에만 진행합니다.
-
----
-
-## 8. R011과의 관계
-
-Phase 1 COEXIST 기간 중 R011 (SHOULD-memory-integration) 적용:
-
-| R011 항목 | COEXIST 수정 사항 |
-|-----------|-----------------|
-| Primary: Native auto memory | 변경 없음 |
-| Supplementary: claude-mem | 변경 없음 (계속 사용) |
-| Session-End Self-Check | claude-mem + AgentMemory 둘 다 저장 시도 |
-| Failure Policy | 둘 중 하나 실패해도 비차단 |
-
-세션 종료 시 자가 점검 (COEXIST 확장):
-
-```
-1. sys-memory-keeper가 MEMORY.md 갱신? → YES
-2. claude-mem 저장 시도? → YES (기존)
-3. AgentMemory 저장 시도? → YES (COEXIST 추가)
-모두 완료 후 사용자에게 확인
-```
-
----
-
-## 9. 참고
-
-- **이슈**: #1169 (AgentMemory 마이그레이션 계획)
-- **이전 단계**: [measure-step-zero.md](./measure-step-zero.md)
-- **R011**: `.claude/rules/SHOULD-memory-integration.md` (Dual-Backend Advisory 섹션)
-- **관련 기억**: [[project-sequencing-alpha-beta-gamma]]
-- **자산 처리표**: `guides/agentmemory-migration/asset-disposition.md` (Phase 2 전 검토 필요)
-- **measure 스크립트**: `scripts/measure-claude-mem-usage.sh`
-- **롤백 가이드**: #1169 본문 조치 4 (30분 롤백 절차)
+Keep this guide as a short historical pointer until downstream documentation no longer links to the old AgentMemory migration path.

package/templates/guides/index.yaml

@@ -107,7 +107,7 @@ guides:
       type: internal
 
   - name: agentmemory-migration
-    description: AgentMemory migration measurement and COEXIST rollout guide for replacing legacy claude-mem
+    description: Retired external memory migration guide documenting native-memory-only policy and approved searchable backend usage
     path: ./agentmemory-migration/
     source:
       type: internal

package/templates/guides/multi-agent-debate-patterns/README.md

@@ -4,7 +4,7 @@
 
 | Pattern | Goal | Use When |
 |---------|------|----------|
-| `agora` | Reach adversarial consensus | Release gates, design approval, high-risk specs |
+| `roundtable-debate` | Preserve dissent while reaching a bounded decision | Release gates, design approval, high-risk specs with minority risks |
 | `roundtable-debate` | Preserve dissent | Strategy choices, tradeoffs, ambiguous product or architecture decisions |
 
 ## Failure Modes

package/templates/guides/multi-provider-exec/README.md

@@ -1,83 +1,13 @@
-# Multi-Provider Exec
+# External Interop Guidance
 
-## Overview
+The packaged multi-provider exec skills have been retired. For Codex interoperability, use the official Claude Code plugin `openai/codex-plugin-cc` only when it is explicitly installed and requested.
 
-Unified reference for executing prompts through external LLM providers via exec skills. Complements the [Multi-Model Routing](../multi-model-routing/README.md) guide (Claude model selection) with cross-provider execution capabilities.
+## Current Paths
 
-Inspired by OpenHarness's provider profile switching pattern, adapted for oh-my-customcodex's skill-based architecture.
+| Need | Preferred path | Notes |
+|------|----------------|-------|
+| Codex interop | `openai/codex-plugin-cc` | Official plugin path; opt-in only. |
+| Token-optimized local command output | `rtk-exec` | Existing RTK proxy remains supported. |
+| Research or independent review | `researcher`, expert agents, or `roundtable-debate` | Prefer in-repo agent workflows unless plugin interop is requested. |
 
-## Provider Matrix
-
-| Provider | Skill | CLI Dependency | Model | Strengths |
-|----------|-------|---------------|-------|-----------|
-| OpenAI (Codex) | `codex-exec` | `codex` CLI | GPT-5.4 | Code generation, broad knowledge |
-| Google (Gemini) | `gemini-exec` | `gemini` CLI | Gemini 2.5 Pro | Long context, multimodal |
-| RTK (proxy) | `rtk-exec` | `rtk` CLI | Configurable | Token-optimized output, cost reduction |
-
-## Availability Detection
-
-The `session-env-check.sh` hook (SessionStart) auto-detects available providers:
-
-```
-[SessionStart] Checking external CLI availability...
-  codex: ✓ available
-  gemini: ✗ not found
-  rtk: ✓ available
-```
-
-Providers are opt-in — missing CLIs are silently skipped.
-
-## Usage Patterns
-
-### Direct Invocation
-
-```
-/codex-exec "Review this function for security issues"
-/gemini-exec "Analyze this architecture diagram"
-/rtk-exec "List files matching pattern X"
-```
-
-### Provider Selection Guide
-
-| Task | Recommended Provider | Rationale |
-|------|---------------------|-----------|
-| Second opinion on code review | codex-exec | Independent model reduces confirmation bias |
-| Long document analysis | gemini-exec | 1M+ context window |
-| Token-heavy batch operations | rtk-exec | Compressed output reduces context cost |
-| Security audit cross-check | codex-exec | Different training data catches different patterns |
-| Multi-model verification | All three | `/multi-model-verification` skill orchestrates this |
-
-### Integration with Existing Skills
-
-| Skill | Uses Provider | How |
-|-------|--------------|-----|
-| `multi-model-verification` | codex-exec + gemini-exec | Parallel verification with severity classification |
-| `reasoning-sandwich` | Any exec skill | Pre/post reasoning with different models |
-| `model-escalation` | Claude models only | Internal escalation (haiku→sonnet→opus), not cross-provider |
-
-## Relationship to Multi-Model Routing
-
-| Aspect | Multi-Model Routing | Multi-Provider Exec |
-|--------|--------------------|--------------------|
-| Scope | Claude model selection | Cross-provider execution |
-| Models | haiku / sonnet / opus | GPT-5.4 / Gemini 2.5 / RTK proxy |
-| Mechanism | `model` frontmatter field | Exec skill invocation |
-| Use case | Cost/quality optimization within Claude | Independent verification, specialized tasks |
-| Guide | `guides/multi-model-routing/` | `guides/multi-provider-exec/` |
-
-## Configuration
-
-No global configuration required. Each exec skill reads its own CLI configuration:
-
-| Skill | Config Source |
-|-------|-------------|
-| codex-exec | `~/.codex/config` or CODEX_API_KEY env |
-| gemini-exec | `~/.gemini/config` or GEMINI_API_KEY env |
-| rtk-exec | RTK proxy running on localhost |
-
-## Limitations
-
-- Provider availability depends on user's CLI installations
-- Cross-provider results are advisory — Claude remains the primary execution engine
-- No automatic fallback between providers (by design — explicit selection preferred)
-- Rate limits and costs are provider-specific and not tracked by oh-my-customcodex
+Do not auto-delegate to retired provider wrapper skills. Keep expert agents responsible for reviewing any plugin-assisted output.

package/templates/manifest.json

@@ -1,11 +1,11 @@
 {
-  "version": "0.5.7",
+  "version": "0.5.9",
   "requiresCC": ">=2.1.121",
   "claudeCode": {
     "minimumVersion": "2.1.121",
     "protectedPathBypassVersion": "2.1.126"
   },
-  "lastUpdated": "2026-05-29T00:00:00.000Z",
+  "lastUpdated": "2026-05-30T00:00:00.000Z",
   "components": [
     {
       "name": "rules",
@@ -23,7 +23,7 @@
       "name": "skills",
       "path": ".agents/skills",
       "description": "Reusable skill modules (project-scoped repo skills)",
-      "files": 123
+      "files": 120
     },
     {
       "name": "guides",

package/templates/.claude/skills/agora/SKILL.md

@@ -1,209 +0,0 @@
----
-name: omcustomcodex:agora
-description: "Multi-LLM adversarial consensus loop — 3+ LLMs compete to find flaws in designs/specs until unanimous agreement is reached"
-user-invocable: true
-argument-hint: "<document-path> [--rounds N] [--severity-threshold HIGH]"
-effort: max
-scope: core
-version: 1.0.0
-source:
-  type: external
-  origin: github
-  url: https://github.com/baekenough/baekenough-skills
-  version: 1.0.0
----
-
-# Agora: Multi-LLM Adversarial Consensus
-
-3개 이상의 LLM(Claude, Codex/GPT, Gemini)이 경쟁적으로 설계/문서의 결함을 찾고, 만장일치 합의에 도달할 때까지 반복하는 적대적 교차 검증 스킬.
-
-## Prerequisites
-
-- `codex-exec` skill (Codex/GPT 호출)
-- `gemini-exec` skill (Gemini 호출)
-- Agent Teams enabled (`OMCODEX_AGENT_TEAMS=1`) or Agent tool available
-
-## Sensitive-Path Delegation
-
-Sensitive-path compatibility note: if this skill delegates work that touches `.claude/**`, `.claude/outputs/**`, `templates/.claude/**`, or read-only measurements of those paths, keep `.codex/**` edits on the normal Codex path. On Claude Code v2.1.121+ with `bypassPermissions`, direct writes to `.claude/skills/`, `.claude/agents/`, and `.claude/commands/` are allowed; on v2.1.126+ that extends to broader protected paths. Only use `/tmp/{skill}-{timestamp}.md` as a legacy fallback when the target runtime is older or still prompts.
-
-## Usage
-
-```
-/omcustomcodex:agora docs/design.md                           # Default: 3 LLMs, unlimited rounds
-/omcustomcodex:agora docs/design.md --rounds 10               # Max 10 rounds
-/omcustomcodex:agora docs/design.md --severity-threshold HIGH # Exit when no HIGH+ findings
-/omcustomcodex:agora docs/design.md --models claude,codex     # 2 LLMs only
-```
-
-## Workflow
-
-### Phase 1: Setup
-1. Read the target document
-2. Create Agent Team: `TeamCreate("agora-review")`
-3. Create review tasks per focus area
-
-### Phase 2: Spawn Reviewers (parallel)
-Spawn 3 reviewers as Agent Team members:
-
-```
-
-### Anti-Groupthink Mode
-
-Use `--anti-groupthink` when consensus itself is a risk:
-
-1. Reviewers submit independent findings before seeing peer output.
-2. One reviewer is assigned as devil's advocate.
-3. Minority findings are preserved unless the synthesis explicitly rejects them with evidence.
-4. Debate is capped at two challenge rounds before the lead either decides or requests more facts.
-
-For decisions where dissent preservation is the main goal, use `roundtable-debate` directly instead of `agora`.
-Agent(name: "claude-critic", model: opus, effort: max)
-  → 20-point deep adversarial review
-  
-Agent(name: "codex-critic", model: opus)
-  → Invoke Skill(codex-exec) for GPT perspective + independent Claude analysis
-  
-Agent(name: "gemini-critic", model: opus)  
-  → Invoke Skill(gemini-exec) for Gemini perspective + independent Claude analysis
-```
-
-### Phase 3: Independent Review
-Each reviewer performs adversarial review with this template:
-
-```
-For EACH review point:
-### Round N: [Topic]
-**Severity**: CRITICAL / HIGH / MEDIUM / LOW
-**Flaw**: [Specific, concrete problem description]
-**Evidence**: [Why this is real, not theoretical]
-**Impact**: [What happens if not addressed]
-**Counter-argument**: [Best case FOR the current design]
-**Verdict**: KEEP / MODIFY / REJECT
-```
-
-Review areas (adapt to document type):
-- Architecture fundamentals
-- Component/service design
-- Data architecture
-- Security & resilience
-- Feasibility & deployment
-- Testing strategy
-- Operational complexity
-
-### Phase 4: Cross-Review (Peer-to-Peer)
-Each reviewer sends findings to the other two via `SendMessage`.
-
-Counter-review template:
-1. Which findings do you **AGREE** with? (and why)
-2. Which findings do you **DISAGREE** with? (evidence-based rebuttal)
-3. What did they **MISS** that you caught?
-4. What did they catch that you **MISSED**?
-5. **SEVERITY** adjustments — upgrade or downgrade with justification
-
-### Phase 5: Synthesis
-Team lead aggregates all findings:
-
-```
-UNANIMOUS CRITICAL: [findings all 3 agreed on]
-STRONG AGREEMENT:   [findings 2/3 agreed on]
-SPLIT DECISIONS:    [findings with disagreement + resolution]
-```
-
-Determine verdict:
-- **BUILD**: No CRITICAL, no unresolved HIGH
-- **BUILD WITH CHANGES**: No CRITICAL, HIGH findings have accepted mitigations
-- **REDESIGN**: Any unresolved CRITICAL findings
-- **ABANDON**: Fundamental concept is flawed
-
-### Phase 6: Loop (if REDESIGN)
-1. Team lead produces/delegates redesign addressing ALL critical findings
-2. New version sent to ALL reviewers: `SendMessage(to: "*")`
-3. Reviewers re-review → GOTO Phase 4
-4. Repeat until EXIT criteria met
-
-### Phase 7: Exit (consensus reached)
-When ALL reviewers agree BUILD or BUILD WITH CHANGES:
-1. Produce final consensus report
-2. Write to `.codex/outputs/sessions/{date}/agora-{topic}-{time}.md`
-3. Shut down team: `SendMessage(to: "*", message: {type: "shutdown_request"})`
-
-## Reviewer Principles
-
-1. **NEUTRAL** — no reviewer has home team advantage
-2. **COMPETITIVE** — find flaws others missed
-3. **CRITICAL** — "fewer than 5 CRITICAL flaws = not looking hard enough"
-4. **EVIDENCE-BASED** — every finding cites specific evidence
-5. **CONSTRUCTIVE** — every flaw includes recommended fix
-6. **CONVERGENT** — goal is consensus, not endless disagreement
-
-## Consensus Criteria
-
-| Condition | Required |
-|-----------|----------|
-| CRITICAL findings resolved | ALL |
-| HIGH findings resolved or accepted | ALL |
-| All reviewers rate BUILD or BUILD WITH CHANGES | YES |
-| Cross-review disagreements resolved | ALL |
-
-## Output Format
-
-```markdown
-# Agora Consensus Report
-
-## Document: [path]
-## Rounds: [N]
-## Reviewers: [list with LLM models used]
-
-## Verdict: [BUILD / BUILD WITH CHANGES / REDESIGN]
-
-## Unanimous Findings
-| # | Finding | Severity | All 3 Agree |
-|---|---------|----------|-------------|
-
-## Required Changes Before Build
-1. [change with source reviewer]
-2. ...
-
-## Accepted Risks
-- [finding accepted with justification]
-
-## Unique Contributions Per Reviewer
-| Reviewer | Findings Others Missed |
-|----------|----------------------|
-
-## Process Metrics
-- Rounds: N
-- Total findings: N
-- Cross-adopted: N
-- Severity upgrades: N
-- Severity downgrades: N
-- Disagreements raised: N
-- Disagreements resolved: N/N
-```
-
-## Configuration
-
-```yaml
-# Default settings
-agora:
-  max_rounds: unlimited       # Set --rounds to limit
-  severity_threshold: HIGH    # EXIT when no findings >= threshold
-  models:
-    - claude (opus, max effort)
-    - codex (via codex-exec skill)
-    - gemini (via gemini-exec skill)
-  review_points: 20           # Per reviewer
-  cross_review: true          # Peer-to-peer sharing
-  auto_redesign: true         # Auto-produce redesign on REDESIGN verdict
-```
-
-## Anti-Patterns
-
-| Anti-Pattern | Why Wrong | Correct |
-|-------------|-----------|---------|
-| Single LLM review | Misses blind spots | 3+ LLMs find complementary flaws |
-| No cross-review | Reviewers don't challenge each other | Peer-to-peer sharing surfaces disagreements |
-| Accepting first BUILD | May miss edge cases | Loop until ALL agree |
-| Ignoring split decisions | Unresolved disagreements fester | Resolve every split with evidence |
-| Push for consensus too fast | Premature agreement | Let reviewers challenge freely |