Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/Minokun/xdevnpx agentmods add skills/minokun/xdev/xdev-researchWrote 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/minokun/xdev/xdev-research)<a href="https://agentmods.dev/skills/minokun/xdev/xdev-research"><img src="https://agentmods.dev/badge/skills/minokun/xdev/xdev-research/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/minokun/xdev/xdev-research"><img src="https://agentmods.dev/badge/skills/minokun/xdev/xdev-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Rogue Agent · line 171 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
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 | $0.00008 | $0.11072 |
| Opus 5.5 | $0.00003 | $0.04429 |
| Sonnet 5.5 | $0.00002 | $0.02214 |
| Haiku 4.5 | $0.00001 | $0.01107 |
Grade A, and why
xdev-research 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 16d 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 — 418 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/xdev-research — 算法研究流程
研究输入： 用户消息（手势之外的原文）
本流程是 xdev 的研究特化——复用 full-dev 的整套机制，只替换审查维度：面板 + 门下门（A1–A4 → R1a–R1c + R1）、worker receipt、 drift check（基准换成预注册 proposal.md）、对抗审查（附录 D → 报告对抗审查）。映射：对照基准 = 预注册的假设与判定阈值 （≈ Intent Contract），真实执行 = 实验真跑且指标落盘，fresh 审核 = 研究视角的审查员面板 + 结果溯源审计。 硬规则 1–5 全部生效（定义见
/xdev-full-dev，速览见本文附录 H），另有研究专属铁律 T1–T3（见下）。 设计依据（均已在线验证）：自动实验（AI-Scientist / Agent Laboratory / RE-Bench / SWE-bench）、DeepResearch（STORM、 multi-agent research：三视角开局 + 文件系统产物指针 + replan 显式退出）；OSF 预注册、晶体学 R-free、cryo-EM 金标准半图、 CASP 机器盲判、wwPDB OneDep 归档、FLT/Prove2Me 先冻结后执行、NCI append-only 哈希链——出处标在对应条款旁。
研究专属铁律（在通用 5 条之上，违反即流程失败）
- T1 预注册不可逆：假设（H0/H1）、判定阈值、评测集、seed 数一经生效——生效路径两条（显式并列）：
路径 A：阶段 2 门禁 approve 后经用户单次确认；路径 B：已签授权包的轮次，门下门 approve 即生效
（用户确认已在授权包签署时前置完成）——分析阶段只许判定不许改；
要改 = 显式回到阶段 2 重新过门禁并留痕。
T1 的事实性前提例外（与
/xdev-full-dev阶段 1 的外部接地检查同构）：预注册里凡出现 "这个数据拿不到 / 只能手工构造 / 没有现成实现 / 上游不支持"这类关于外部世界的断言， 它不是"已批准的决策"，而是待核实的断言——必须先验证（取回权威源 / 跑通一次最小复现）， 或显式标注"未验证"并说明它若为假会怎样推翻整个方向。 理由：预注册是更强版本的设计公理，而开发侧的实盘已经证明——一个被写成前提的错误断言 会让全部下游闸门（12 位审核员 × 6 轮）在结构上被禁止质疑它。研究侧的代价更大：不是返工一次交付， 而是烧掉一整轮实验预算后才发现方向建在假前提上。 写法：前提：<断言> | 证据：<命令输出/来源 URL/file:line> | 或：未验证→它会推翻什么。 - T2 数字必须溯源：报告/结论中的每个数字必须指向
runs/下真实存在且通过哈希链校验的 metrics 文件路径；由模型凭记忆写出的数字一律无效。评测集固定，禁止子采样挑好结果。 溯源目标必须在持久位置（研究目录/仓库入库）；指向 /tmp、会话缓存等易失位置的证据视为无证据。 解析脚本自身也要可伪证（硬规则 2 的研究侧形态）：喂一份故意改坏的日志/指标，解析脚本 必须报错或算出不同的数——恒绿的解析脚本与恒绿的断言等价，而它会污染整条溯源链。 - T3 负结果同样落盘：假设被证伪、实验失败、方向走不通，都必须写入报告与实验记录， 不得只报喜。用户喊停只阻止开新轮，不豁免已产生数据的报告义务。
- 失误复发 → 机械化（同
/xdev-full-dev附录 E）：执行侧同类失误第二次出现 → 写进 run-check / 解析脚本或停止条件；只加文字规则 = 未闭环（handoff.mjs failure-check红）。
目录约定（默认值，按项目惯例可调）
docs/research/<date>-<slug>/ # node ../../bin/research.mjs init <dir> 一键 scaffold（路径相对本 skill 资源目录，同 full-dev 阶段 4）
literature.md # 文献与现状扫描（阶段 1）
directions.md # 方向看板：候选方向清单 + 每轮后的状态更新（阶段 1 建，轮级循环维护）
# 只记方向级结论（证实/证伪/搁置+证据），不记 run 状态
proposal.md # 假设 + 实验方案 + 预注册验收（阶段 2 产出并冻结；补跑追加为 append-only）
matrix.yaml # 机器可读实验矩阵（阶段 2 随方案冻结 = "陈述"，阶段 3 执行）
judge/ # judge 判定脚本 + 解析脚本 + fixtures + 单测 + 冻结内容哈希
runs/<run-id>/ # 每 run 独立目录：config 副本 + append-only 日志 + metrics.json
# + runtime.json（pid/job 句柄）+ receipt.md + manifest.json（溯源声明，run-check 校验）
runs/VOIDED.md # 作废 run 名单（research.mjs void 登记，只增不删；作废 run 不得被引用）
report.md # 报告，按轮追加；终稿含结局核对表与 judge 原始输出（阶段 5）
state.md # 跨会话状态：固定头部（当前阶段/已过门禁/待办/活跃 job）+ run 状态 +
# 哈希链清单。只记 run 级状态，不记方向结论（与 directions.md 互斥）
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.
- 16d ago Changed 46276bdda34d
- 1mo ago Changed · +9 lines c5e677874955
- 1mo ago Changed · +26 lines e5039f0c0e04
- 1mo ago Changed · +16 lines 39b60c683b62
- 1mo ago Changed · +34 lines e9d23db7d369
- 1mo ago First seen · 333 lines · 8 tokens per session scan A e8246bc80fac
xdev-research is a skill published in the GitHub repository Minokun/xdev (24 stars, last pushed yesterday), licensed MIT. It adds 8 tokens to every session and 11,072 once invoked, about $0.0000 per session on Opus 5.5. 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 2026-09-06.
Other skills, from other repositories
instrument-data-to-allotrope
Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full…
exploratory-data-analysis
Performs bounded, local exploratory analysis of explicitly supported scientific files. Supports redacted CSV/TSV/JSON profiles; optional NumPy, HDF5, FASTA/FASTQ, and basic image metadata inspection; missingness/leakage audits; outlier and transformation sensitivity; and rigorous EDA report scaffolds. Other domain…
matlab
Builds, reviews, migrates, and plans MATLAB or GNU Octave numerical workflows. Use for arrays, tabular/time data, tests, projects, graphics, MAT files, and explicit Python interoperability.
phylogenetics
Builds and analyzes phylogenetic trees using MAFFT multiple sequence alignment, IQ-TREE maximum likelihood with ModelFinder and branch support, and FastTree approximate inference. Uses ETE3 for tree summaries and visualization. Applies to homologous nucleotide or protein sequences, microbial gene trees, protein…
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
presenting-conference-talks
Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script. Use when preparing oral talks, spotlight presentations, or invited talks for ML and systems conferences.