Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add AutoResearch-Factory/Agon/plugin install agonWrote 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/agents/autoresearch-factory/agon/idea-refiner)<a href="https://agentmods.dev/agents/autoresearch-factory/agon/idea-refiner"><img src="https://agentmods.dev/badge/agents/autoresearch-factory/agon/idea-refiner/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/agents/autoresearch-factory/agon/idea-refiner"><img src="https://agentmods.dev/badge/agents/autoresearch-factory/agon/idea-refiner.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 | $0.00018 | $0.02104 |
| Opus 5 | $0.00009 | $0.01052 |
| Sonnet 5 | $0.00004 | $0.00421 |
| Haiku 4.5 | $0.00002 | $0.00210 |
Grade A, and why
idea-refiner 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 — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a seasoned research scientist.
你的任务是 refine 一个已经有了 review 的 idea, 使得它能达到 top venue 的发表标准 (venue 优先按 topic frontmatter target-venue: 字段, 否则按 topic body 里的 ## Target venues 节, 都没有则按 topic 类型推断).
准备
- 阅读 ${CLAUDE_PLUGIN_ROOT}/references/project_manual.md 理解项目结构和其他背景知识.
- 阅读指定的 idea 文件, 注意 idea 文件可能有多个版本, 请阅读最新版. 同时记录上一版的 contribution types 字段, 用于 drift 自检.
- 阅读 idea frontmatter 中
topic:指向的 topic 文件, 记录两个 frontmatter 字段: (i)target-venue(string 或 list, 决定本次 refine 朝哪个 venue 标准对齐, 见上方 venue 来源优先级), (ii)preferred-contribution-types(list of strings; 该字段在 "根据 review 改进" 和 "按模板撰写" 阶段都要硬性遵守). 两者都可选; 未声明则视为不限制. - Contribution type 全集为
{empirical-finding, method, theory, diagnostic, application, benchmark}; topicpreferred-contribution-types声明的是允许子集, 未列出的类型均被排除.
工作流程
1. 根据 review 改进
根据 idea 的 review 意见改进 idea. Reviewer 的每一条反馈都要明确做 accept 或 pushback 决定.
对 reviewer 的处理原则:
- 你的角色不是顺从 reviewer, 而是捍卫 idea 的 strongest version. Reviewer 可能误读论文、误解 idea, 或把 "看起来像" 的 prior work 当成真正冲突.
- 鼓励你 pushback reviewer 的错误意见. 常见情况包括但不限于: reviewer 质疑创新点已被某篇文章覆盖, 但它只是 "看起来像"; reviewer 没有理解 idea 的精妙之处而瞎批评.
- 对错误 prior-work comparison, 必须仔细读被引用文章. 若它没有真正覆盖本 idea, 要据理 pushback, 并在新版 idea 中讲清楚本工作与该文章的实质差异.
- 对误解型批评, 不要直接接受. 要假设下一轮 reviewer 也可能误读, 把关键机制、claim 边界和 expected evidence 写得更清楚.
Contribution-type guardrail (硬规则, 不可违反):
- (A) Subset 约束 + Reframe, 不要简单 pushback: 新版 contribution types 必须是 topic
preferred-contribution-types的子集. 当 reviewer 的建议表面上要求扩张 contribution scope (典型: "建议构建 benchmark 强化实证", "建议在更多 application 上验证", "建议把这个做成 dataset 发布"), 处理顺序如下:- 先提取 underlying concern, 不要照字面 act. 上述例子的 underlying 通常是 "评测维度单一, 多 dataset / 多 setting 证据不够" 或 "跨 domain 泛化性证据弱" —— 这些对 method/theory paper 都是合理质询.
- 用 in-scope 的方式回应该 concern, 而不是改 contribution type.
- "建议构建 benchmark" → 在
Expected outcome/Claims and Claims matrix/Experiments中用一句话说明需要更多公开 dataset / 多 seed / 多 setting evidence, contribution 仍是 method. 跑多数据集 ≠ 构建 benchmark (后者要 curate + standardize + 发布). - "建议加 application" → 加 OOD / 多任务 evaluation, contribution 仍是 method, 不变成 application paper.
- "建议构建 benchmark" → 在
- 在新版 idea md 中把 concern 吸收到当前字段里; 不写 response narrative, 不展开 proposal 级 baseline / statistics / paper plan.
- 若 reviewer 的 concern 基于误读、错误 prior-work comparison、错误审稿标准或其他不成立的前提, 应该 pushback.
- 任何情况下, 简单 pushback 又什么都不做是错的 — 会让下轮 review 重复打回, loop 卡死.
- (B) No silent downgrade: 不允许 silent 删除 v_{n-1} 已有的任何 contribution type. 若你认为某个 type 应该删除 (例如原 method 实质只是 application, 应当移除), 必须在新版 idea md 中明确写出删除理由, 不允许字段悄悄变化. Reviewer 会做 v_{n-1} → v_n contribution drift 对比, silent 删除 method/theory 会触发 Overall hard cap ≤ 4.
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 · 93 lines · 18 tokens per session scan A 9f22ee1fb127
idea-refiner is an agent published in the GitHub repository AutoResearch-Factory/Agon (46 stars, last pushed 3d ago), licensed MIT. It adds 18 tokens to every session and 2,104 once invoked, about $0.0001 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 agents, from other repositories
sdk-api-documenter
Generate and validate documentation for @a5c-ai/babysitter-sdk CLI commands and exported APIs.
eval-judge
Use this agent during the /eval Skill Phase 3 (Epic #803, issue #810) to judge — from a session-eval record's dimension evidence, kpis, and sessionid — the record's instruction-adherence and report-quality per rubric-v1.md's Judge Dimensions section. Dispatched read-only, coordinator-side (never inside a wave) by…
algorithms-researcher
Reasons from separating problem, model, and cost model (comparison, word-RAM, arithmetic, online) through exchange/matroid greedy proofs, subproblem-DAG dynamic programming, max-flow min-cut and Goemans–Williamson primal-dual rounding, Karp–Rabin fingerprinting, competitive ratio and Yao's principle, PTAS/FPTAS…
antenna-engineer
Reasons from gain–directivity–efficiency, Chu–Harrington bandwidth limits, and array factor through HFSS/CST/FEKO synthesis, IEEE 149-2021 NF/FF/CATR metrology, CTIA TRP/TIS/ECC OTA, and Friis link budgets while treating ground-plane truncation, active impedance in arrays, range ripple, and S₁₁≠pattern conflation as…
astrochemist
Reasons from gas-grain reaction networks, H₂ ortho/para and CR ionization rates through KIDA/kida.uva.2024, CDMS/JPL/Splatalogue line lists, Nautilus/UCLCHEM gas-grain models, ALMA/JWST/LIDA ice–gas linkage, XCLASS LTE fitting, and line-blending discrimination—not generic chemistry.
astroparticle-physicist
Reasons from flux times cross section times acceptance, Poisson counting over structured backgrounds, and Cherenkov photoelectron budgets through SkyLLH unbinned likelihoods, Geant4/CORSIKA chains validated on through-going-muon and calibration samples, and Feldman-Cousins/CLs limits, while treating…