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 skills add Eivor79/ai-debate-plugin --skill ai-debategit clone --depth 1 https://github.com/Eivor79/ai-debate-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/eivor79/ai-debate-plugin/ai-debate)<a href="https://agentmods.dev/skills/eivor79/ai-debate-plugin/ai-debate"><img src="https://agentmods.dev/badge/skills/eivor79/ai-debate-plugin/ai-debate/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/eivor79/ai-debate-plugin/ai-debate"><img src="https://agentmods.dev/badge/skills/eivor79/ai-debate-plugin/ai-debate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00132 | $0.02124 |
| Opus 5 | $0.00066 | $0.01062 |
| Sonnet 5 | $0.00026 | $0.00425 |
| Haiku 4.5 | $0.00013 | $0.00212 |
Grade A, and why
ai-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 9d 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 — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Debate Review workflow
The review workspace is a file-based, multi-agent debate space: agents (Claude, Codex) exchange
design → attack → rebuttal → decision documents about one topic, adversarially verifying each other's
claims, and converge on a decision delivered to the user. It is NOT a finished-docs store; confirmed
knowledge graduates to the wiki, in-flight argument stays in the workspace. Default workspace folder:
llm_wiki/ai_debate/ (configurable; the coordinator is folder-name-agnostic).
Default happy path — one command, hands-free to the verdict
When the user names something to debate/review ("X를 토론해봐", "debate whether we should X",
"리뷰 붙여줘"), the ENTIRE flow is one command — /review-new <topic> — which:
- scaffolds the workspace if missing (no separate init step),
- creates the topic in one shot — slug derived from the user's phrase,
topic.mdwritten from conversation context (do not interrogate the user; one clarifying question max),auto=true, - starts the coordinator in the background (
run_auto.ps1 -Watch; the single-instance mutex makes a duplicate start harmless), so the agents cycle design → attack → rebuttal → decision by themselves, - waits in the background (
wait_for_review.ps1 <topic> -UntilStatusLike decided*,run_in_background) and, on completion, reports thedecision.mdverdict: adopted findings, ruling, residual risks, next step.
Do NOT hand-write rounds yourself, do NOT poll. Multiple topics? Run /review-new for each; the queue
drains by priority. A round cap (default 5 numbered docs, per-topic max_rounds) forces a JUDGE verdict
if the debate ping-pongs, so autonomous runs always terminate.
Round count control — "이 주제 5라운드로" applies immediately
1 round = 1 numbered doc (design = round 1, attack = round 2, rebuttal = round 3, …). "5 rounds"
means 5 numbered docs, then a forced JUDGE writes decision.md.
- At creation — user says "5라운드로 토론해줘" / "debate this in 5 rounds": pass
--rounds 5to/review-new(setsmax_rounds: 5in the topic'sstatus.json). - Existing / already-running topic — user says "해당 주제 5라운드로 해줘" / "make that topic 5 rounds":
update the topic's
status.jsonimmediately via the workspace script:update_status.ps1 -TopicDir <topic-path> -Set @{ max_rounds = 5 } -ForceThis takes effect on the coordinator's very next poll — no restart needed (the cap is re-read fromstatus.jsonevery turn). Confirm to the user: current numbered-doc count vs the new cap (if the topic already has ≥ N docs, the next turn becomes the JUDGE round). max_rounds: 0(or absent) = coordinator default (-MaxNumberedDocs, 5).
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.
- 9d ago First seen · 106 lines · 132 tokens per session scan A db69c33037b9
ai-debate is a skill published in the GitHub repository Eivor79/ai-debate-plugin (1 stars, last pushed 2mo ago), licensed MIT. It adds 132 tokens to every session and 2,124 once invoked, about $0.0007 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 2026-08-31.
Other skills, from other repositories
Test Assertion Effectiveness 常に PASS するテストの検出
A test review that finds assertions—checks in tests—that do not really verify the program under test and therefore may always pass.
omnicheck-gitlab
Use when checking if MR review findings have been applied — verifies both OmniForge-generated and human reviewer comments against the current diff, posts nudge replies on unaddressed threads.
Knowledge-to-Code Alignment 新知識のコード反映・設計知識の保全
A code review check for whether new requirements and domain knowledge are reflected in names, responsibilities, and module boundaries without losing important design history.
Review Criteria Integrity レビュー基準・品質ゲートの自己弱体化検出
A review check for changes that weaken the rules or quality gates used to review the same change.
omnicheck-github
Use when checking if PR review findings have been applied — verifies both OmniForge-generated and human reviewer comments against the current diff, posts nudge replies on unaddressed threads.
omnicreate-gitlab
Use when creating a GitLab merge request (OmniForge). Auto-populates title and description from commits, supports draft MRs, labels, assignees, reviewers, and issue linking.