Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/ggombee/code-forge/debatenpx skills add ggombee/code-forge --skill debategit clone --depth 1 https://github.com/ggombee/code-forgeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00072 | $0.02450 |
| Opus 5 | $0.00036 | $0.01225 |
| Sonnet 5 | $0.00014 | $0.00490 |
| Haiku 4.5 | $0.00007 | $0.00245 |
Grade A, and why
debate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 335 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debate Skill
중요 기술 결정 시 다각도 검증을 위한 교차 모델 토론 스킬. 3가지 모드 중 환경에 맞는 방식을 자동 선택하여 최대 3라운드 진행 후 합의 도출.
목적
| 목적 | 설명 |
|---|---|
| 다각도 검증 | 단일 관점의 맹점 제거 |
| 설계 결정 강화 | 아키텍처 선택의 근거 명확화 |
| 리스크 사전 발견 | 반론을 통한 잠재 문제 탐지 |
| 팀 합의 촉진 | 토론 요약으로 의사결정 공유 |
모드 (3가지)
Mode 1: agent-teams (최상위)
Agent Teams로 팀원을 spawn하여 실시간 병렬 토론. Claude Max + CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 필요.
TeamCreate("debate-team")
→ Agent("advocate", team_name="debate-team") # 찬성
→ Agent("critic", team_name="debate-team") # 반대/비평
→ codex exec (Bash) # Codex 외부 관점
↓
라운드 1~3: SendMessage로 팀원 간 공방
↓
팀 리더(Claude)가 합의 도출
↓
TeamDelete("debate-team")
팀 구성:
| 팀원 | 역할 | 모델 | subagent_type |
|---|---|---|---|
| advocate | 입장 A 지지, 근거 제시 | sonnet | general-purpose 또는 code-forge:architect |
| critic | 입장 B 반론, 가혹한 비평 | sonnet | code-forge:critic |
| codex | 외부 모델 관점 (Bash로 실행) | gpt-5.4 | Codex CLI |
팀 리더 역할 (Claude 본체):
- 주제 정의 및 입장 배분
- 각 라운드 SendMessage로 상대 주장 전달
- 3라운드 후 합의 도출 및 최종 판정
Mode 2: cross-model (Codex CLI)
Agent Teams 없이 Claude + Codex CLI headless 토론.
Claude (입장 A) ↔ Codex CLI (입장 B)
↓
라운드 1: 초기 주장
↓
라운드 2: 반론
↓
라운드 3: 재반론 + 합의 탐색
↓
합의 도출
Codex CLI 실행:
codex exec -s read-only "{프롬프트}"
Mode 3: self-debate (폴백)
Codex 미설정 시 Claude 내부 찬반 토론.
Claude (입장 A: 지지) ↔ Claude (입장 B: 반론)
↓
라운드 진행 (최대 3)
↓
메타 분석: 어떤 주장이 더 강한가?
↓
합의 도출
모드 선택
사용자에게 먼저 묻는다. 자동 선택하지 않는다.
"토론 모드를 선택해주세요:
1. agent-teams — Agent Teams로 팀원 병렬 토론 (Claude Max + 환경변수 필요)
2. cross-model — Codex CLI와 1:1 토론 (Codex 설치 필요)
3. self-debate — Claude 내부 찬반 토론 (별도 설정 불필요)
또는 /debate --mode {모드명}으로 바로 시작할 수 있습니다."
--mode 옵션이 있으면 질문 없이 해당 모드로 즉시 진행.
/debate --mode agent-teams → Mode 1 즉시
/debate --mode cross-model → Mode 2 즉시
/debate --mode self-debate → Mode 3 즉시
선택된 모드의 사전 조건 미충족 시 안내 후 다른 모드 제안.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 335 lines · 72 tokens per session scan A 2644ff5d3d34
debate is a skill published in the GitHub repository ggombee/code-forge (13 stars, last pushed 2mo ago), licensed MIT. It adds 72 tokens to every session and 2,450 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
credit-analysis
固收与信用分析：信用债评级、利差分析、违约风险评估、城投债研究、可转债定价与策略。.
correlation-regime
Correlation-regime detection and crisis attribution — edge-density regime states with hysteresis, causal (no look-ahead) smoothing, regime-aware exposure context, first-mover crisis attribution with honest NAME / MACRO / AMBIGUOUS / ABSTAIN verdicts, and a correlation-rewiring leaderboard that catches slow bleed-outs.
behavioral-finance
Behavioral finance applications: theories of overreaction and underreaction, behavioral explanations for momentum and reversal, investor sentiment cycles, cognitive-bias checklists, and debiasing quantitative strategies.
deep-company-series
Write a publication-grade 8-part deep-dive series on a single company (120k words total): cognitive reset / moat / profit engine / hidden assets / era variable (e.g. AI) / financials Buffett-style / management / valuation+redlines. The core IP is NOT writing but REVISING — a strict fact-check checklist catches…
earnings-revision
Earnings estimate revisions, guidance analysis, and post-earnings drift (PEAD) — track analyst consensus changes, earnings surprise patterns, and management guidance shifts for US/HK equities.
performance-attribution
Performance attribution analysis — Brinson sector/stock-selection attribution, factor alpha/beta decomposition, market-timing evaluation, and benchmark comparison framework.