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 MrSGSA/math-modeling-skill-dify --skill model-reviewergit clone --depth 1 https://github.com/MrSGSA/math-modeling-skill-difyWrote 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/mrsgsa/math-modeling-skill-dify/model-reviewer)<a href="https://agentmods.dev/skills/mrsgsa/math-modeling-skill-dify/model-reviewer"><img src="https://agentmods.dev/badge/skills/mrsgsa/math-modeling-skill-dify/model-reviewer.svg" alt="Measured on agentmods" 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 | $0.00117 | $0.01495 |
| Opus 5 | $0.00059 | $0.00747 |
| Sonnet 5 | $0.00023 | $0.00299 |
| Haiku 4.5 | $0.00012 | $0.00150 |
Grade A, and why
model-reviewer 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 8d 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 — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.
评审手
路径
ROLE_ROOT:本文件所在目录。SKILL_ROOT:ROLE_ROOT/../../..,运行任务时只读。PROJECT_ROOT:用户项目目录,内部审计产物只写这里。
输入
读取题目、附件、建模手产物、全部代码、机器可读结果、复现清单和候选图。已有论文审稿时同时读取论文;评审图片必须逐张实际查看。
内部输入与产物
PROJECT_ROOT/results/result_registry.json:由编程阶段生成,本阶段只校验结构、来源和跨输出一致性。PROJECT_ROOT/results/red_team_audit.json:由本阶段生成。
二者是论文定稿的验证输入,不作为额外最终交付物。用户单独要求复核或审稿时,按主 Skill 额外交付 评审报告.md;只有用户要求修改现有 Word 时才生成不覆盖原稿的 论文修订稿.docx,并按用户选择保留修订或批注。练习复盘文件只在练习、模拟赛或用户明确要求复盘时生成。
执行顺序
- 完整读取
../../评审门与证据等级.md和../../参数语义与约束审计.md,按题型标记每项检查为pass、fail、pending或有理由的na。 - 独立建立主张—证据映射,不沿用原作者的确定性措辞。
- 先核对数据血缘、去重/缺失处理、训练—调参—最终验证隔离、时间或组别泄漏,再攻击核心判据:评价对象、目标函数与题目指标是否对齐、可见域/样本域、充分必要性、量纲、边界、集合并交、时间窗口和极端反例。
- 对所有影响核心结论的拟合参数、决策变量、阈值、锚点和数值松弛量逐项核验语义、单位、允许域、依据、全域行为、可辨识性和输出影响。不得仅检查目标函数、残差、收敛或条件数。
- 参数越界、贴边、反常、跨数据漂移、强补偿或跨模型符号/单调性冲突时,强制运行同数据同预算的无约束—合理约束重拟合;比较核心输出和结论,不得以拟合改善覆盖物理/统计异常。物质性阈值须由题目精度或决策容差预先确定。
- 对多资源方案先把目标统一为“越大越好”的效用口径,再相对无资源基线计算独立效果、联合效果和逐资源删除损失;最小化目标必须反向换算。删除损失为零时不得写成正边际贡献,删除后不可行时单独报告可行性作用,不能用任意无穷大替代。
- 对离散—连续耦合问题记录离散搜索空间、候选筛选、同预算复算和未覆盖范围。只比较过候选时不得声称全局最优。
- 对随机或网格算法执行多种子、扰动和逐级加密;对预测、估计和决策结论检查样本外验证、不确定性校准与指标替代敏感性。阈值由题目精度、数值尺度或预先声明的容差确定,不机械固定为1%。
- 对每个生成核心结论的例程,使用解析解、手算小例、枚举、小规模精确解、不同技术栈独立实现、守恒/不变量或变形测试中的适用项作为 oracle。至少构造一个能区分常见错误实现的反例,检查输入参数是否真正生效、是否混入固定演示答案、目标方向与索引是否正确;待审代码自身及同一公式的重写不能作为独立真值。仅有“程序成功运行”不能证明实现正确。按 1.3 版结构登记
oracle_audit,再校验result_registry.json及摘要、正文、表、图和附录同源。 - 写入
red_team_audit.json,运行:
python scripts/red_team_gate.py "<PROJECT_ROOT>/results/red_team_audit.json"
python scripts/result_registry_gate.py "<PROJECT_ROOT>/results/result_registry.json"
CLI 默认核对所有 evidence 是否指向 PROJECT_ROOT 内实际存在的文件;--schema-only 仅用于测试夹具或审计模板开发,不能作为论文发布依据。
- 重大失败按
reopen_stage回退;修正后只重跑受影响计算和下游产物。评审门通过后才交给论文手。
严重度
critical:判据、数据、约束或实现错误足以推翻核心结论,必须回退。major:结果或最优性证据不足,可能显著改变方案或论文主张,必须修正或降级表述。minor:不改变核心结论的表达、格式或小数精度问题,可在论文阶段修正。
禁止事项
- 不把评审稿、优秀案例或历史复盘当成当前题标准答案。
- 不因测试未执行就填写
pass;条件不适用时写na和理由。 - 不用“多加采样点”“多跑几代”替代对判据和搜索空间的解释。
- 不把可行性、局部稳定性、候选中较优和全局最优混为一谈。
- 不把优化器收敛、残差下降、信息准则改善或更多小数位当作参数合法性的证据。
- 不因异常参数在某些领域可能存在就直接放行;必须给出当前数据域的适用依据或完成约束对照。
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 8d ago First seen · 61 lines · 117 tokens per session scan A e6525725c1fb
model-reviewer is a skill published in the GitHub repository MrSGSA/math-modeling-skill-dify (4 stars, last pushed 1mo ago), licensed MIT. It adds 117 tokens to every session and 1,495 once invoked, about $0.0006 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
math-modeling-solver
A Chinese-language guide for solving mathematical modeling competition problems. It covers China’s CUMCM and America’s MCM/ICM, where teams use mathematics and code to answer real-world problem statements.
math-modeling-paper
A Chinese-language guide for writing papers for mathematical modelling competitions, where teams use mathematics and data to study a real-world problem.
ez-math-model
Use when solving CUMCM, MCM, ICM, or other mathematical modeling tasks that need contest problem intake, model selection, Python solving, figures, paper writing, quality audit, and packaged deliverables.
brainstorming
Use when EZmathmodel model selection is unclear after the modeling decision tree, the problem spans multiple domains, or modeler needs several candidate approaches before writing modelingplan.md.
scientific-slides
Use when the user explicitly asks EZmathmodel to create defense slides, academic presentation slides, Beamer slides, or PPT after the modeling paper has been packaged.
interpret-modeling-problems
A method for turning a mathematical modelling contest problem and its attachments into a checked plan for solving and documenting it.