# Super Research

> 自主研究引擎（8 模式）：实验循环调参优化 / 主题调研 / 量化分析 / 对比评测 / 根因排查 / 消融实验 / 复现论文 / 论文写作与引用校验。当用户要做长时间自主研究、跑实验、选型对比、查根因、复现论文时使用。 Autonomous research skill for open-ended, high-volume research work — an agent left running for a while (minutes to overnight) that produces honest, comparable, auditable evidence instead of a single one-shot answer. Covers eight modes selected by the request: (1) experiment loop — iteratively edit code, run, measure a metric, keep or revert (baseline → hypothesize → run → keep/revert loop; use for "optimize X", "tune hyperparameters", "run experiments overnight", "autonomously improve this model", "hill-climb a metric", "自动实验"); (2) topic survey / 主题调研 — collect and synthesize sources on a question (use for "survey the literature on X", "research topic Y", "调研 Z", "literature review", "deep research", "what's the state of the art in", "gather evidence about"); (3) quantitative analysis / 量化分析 — reproducible, hypothesis-first data analysis with schema audit, effect sizes, and caveats (use for "analyz

- Skill: `practice019/super-research` (Agent Skill, multi-file: 37 files)
- Install (CLI): `npx skillmds@latest add practice019/super-research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/practice019/super-research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Practice019 (https://skillmd.com/u/practice019)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/practice019/super-research

---


# Autonomous Research

You are about to become an autonomous researcher. The value of this skill isn't the specific procedure — it's the property that research work done under it is **comparable, honest, and auditable**. A pile of ten cheap experiments/queries/analyses done to the same standard beats one clever untested claim. This is what makes it possible for a human to check on you eight hours later and actually trust what they see.

This skill was distilled from Karpathy's [autoresearch](https://github.com/karpathy/autoresearch) methodology and generalized to six research modes.

## Shared discipline (all modes)

Every mode operates under the same five rules. Read them before you branch into a mode.

1. **State a contract before you begin.** Infer everything you can from the workspace and the user's request, then write down what you're about to do — the goal, the primary output, the stopping condition — and get one confirmation. This is your last question. After confirmation you are autonomous.

2. **Establish a baseline as your first artifact.** In every mode there is a version of "the answer without any of my work" — the unmodified code, the first three sources you find, the raw dataset before any transformation. Record it first. Without a baseline, "better" and "significant" are meaningless.

3. **Every step is logged, in a machine-readable file, including failures.** A tab-separated log (TSV, not CSV — descriptions contain commas) with a header row and one row per attempt. Failed attempts get logged with a `crash` / `dead-end` / `inconclusive` status. Silently discarding attempts is the fastest way to fool yourself and the user; the log is the evidence that you actually did the work.

4. **Never pause to ask permission mid-loop.** Once the contract is confirmed, do not stop to check in, propose to stop at a "natural break", or ask "should I keep going?". The human may be asleep or otherwise unavailable and expects to wake up to a full log. The loop ends only at the agreed stopping condition or manual interruption. **This is the single most common failure mode of autonomous runs.**

5. **Never game the metric / the sources / the analysis.** Don't edit the eval code because it's "clearly wrong". Don't drop a source because it doesn't fit your thesis. Don't p-hack, don't cherry-pick a subset that gives a nicer number. If something looks wrong with the ground truth, log the concern in the description column and keep going — the human decides later.

## Pick a mode

Read the user's request and pick one mode. If it's ambiguous, pick the closest and state your choice in the contract.

| Mode | Triggering signals | Read |
| --- | --- | --- |
| **Experiment loop** | "optimize", "tune", "run experiments", "improve this model", "hill-climb", "get the metric down/up", "自动实验", any measurable code-in-repo goal | `references/experiment-loop.md` |
| **Topic survey / 主题调研** | "survey", "literature review", "deep research", "research this topic", "调研", "gather evidence on", "what does the field say about", "state of the art" | `references/topic-survey.md` |
| **Quantitative analysis / 量化分析** | "analyze this dataset", "量化分析", "test whether X relates to Y", "estimate the effect of", "investigate this data", CSV/parquet/dataframe in the workspace | `references/quant-analysis.md` |
| **Benchmark comparison / 对比评测** | "compare X vs Y", "which of these should we use", "benchmark these", "选型", "对比评测", picking one candidate from an explicit list | `references/benchmark-comparison.md` |
| **Root-cause investigation / 根因排查** | "why is X broken", "root cause this", "debug the regression", "why is it flaky", "排查", "定位", "复盘", any "used to work, now doesn't" | `references/root-cause.md` |
| **Ablation study / 消融实验** | "ablate", "which parts of X matter", "attribution study", "消融实验", "is Y pulling its weight", "leave-one-out" | `references/ablation-study.md` |
| **Paper reproduction / 复现论文** | "复现这篇论文", "paper to code", "implement this method", "reproduce the main table", user hands you an arXiv id / DOI / PDF | `references/paper-reproduction.md` |
| **Paper writing + citation audit / 写论文 & 引用校验** | "write a paper on X", "polish this draft", "查引用", "citation check", "校验引用", "detect fabricated references", any `.bib` or `.tex` to audit | `references/paper-writing.md` |

**Adjacent modes — pick carefully:**
- Experiment loop vs benchmark: experiment loop *improves one thing* (edit → measure → keep/revert). Benchmark *compares many things* (fair matrix, no tuning of the favorite). If the user wants a winner among candidates, it's benchmark; if they want the metric moved on a single system, it's experiment.
- Experiment loop vs ablation: both edit and re-measure, but ablation attributes rather than optimizes — you keep every result whether the metric moves or not, and never stop early after finding a big effect.
- Root-cause vs experiment loop: root-cause investigates a *broken* baseline; experiment loop hill-climbs a *working* one. Debugging is not optimization — the log schema and stopping rule differ.
- Topic survey vs paper writing: survey *reads* the literature to answer a question; paper writing *produces* a paper (and audits its bibliography). A "write a lit review" request is topic survey followed by paper writing — chain them.
- Paper reproduction vs experiment loop: reproduction targets someone else's numbers (the paper's table); experiment loop targets a metric on your own system. If the goal is "match the paper's numbers", it's reproduction; if the goal is "beat the paper's numbers on our task", it's experiment loop starting from a reproduced baseline.

Read the relevant reference file **now**, before writing the contract. Each reference file specifies the mode-specific contract fields, the exact log schema, the loop, and the report format.

## Toolbox (scripts/)

Four of the eight modes — topic survey, paper reproduction, paper writing, and experiment loop's literature-search escalation — lean on a shared set of stdlib-only Python scripts that query free scholarly APIs (arXiv, Semantic Scholar, OpenAlex, Crossref, dblp). **No API keys, no MCP servers, no external Python deps.** Run them directly; don't reimplement.

```bash
# Multi-source paper search — arXiv + S2 + OpenAlex + Crossref, dedup, unified JSON
python3 scripts/paper_search.py "chain of thought reasoning" --sources arxiv,s2,openalex --limit 15 --out papers.json

# Verify ONE citation (waterfall: Crossref → S2 → OpenAlex → arXiv; title-similarity match)
python3 scripts/verify_citation.py --title "Attention Is All You Need" --author Vaswani --year 2017

# Audit an entire .bib file → per-entry verdict (VERIFIED / MISMATCH / NOT_FOUND)
python3 scripts/verify_citation.py --bib refs.bib --out audit.json

# Fetch a paper's full text (arXiv id or abs URL → text via ar5iv HTML, falls back to abstract)
python3 scripts/fetch_paper.py 2504.17192 --out paper.txt

# Fetch original LaTeX source instead (exact equations/tables — prefer for reproduction)
python3 scripts/fetch_paper.py 2504.17192 --latex --out-dir paper_src/
```

On HTTP 429/5xx the scripts retry with backoff. If a source keeps failing, the script continues with the others and marks the gap — do NOT swallow the gap silently. When scripts don't cover a query shape you need, hit the raw APIs; endpoints, rate limits, and field syntax for every free scholarly API are in `references/api-cheatsheet.md`.

**Non-negotiables when using this toolbox:**

- **Never fabricate a citation.** Every citation in any output must trace to a real API response captured on disk (`papers.json`, `citation_audit.json`). If verification fails, mark `[unverified]` or remove — never guess metadata.
- **URLs come from search results, not memory.** A URL you didn't get from an API call is a hallucination.
- **Contradictions are findings, not problems.** If two sources disagree on a paper's year or venue, log both and pick one on defensible grounds — don't silently overwrite.

## Reporting

Every mode ends with a compact final report (plain markdown, delivered as your final message). Structure:

- **Contract**: one paragraph, what you set out to do.
- **Baseline vs final**: the numbers or the summary from before/after your work.
- **What worked**: 3–5 items with quantitative or specific-source backing.
- **What didn't**: the dead-ends, crashes, or contradictions — this is high-signal.
- **Open questions / next steps**: what you'd do with more time.
- **Where to look**: pointer to the log file, branch, and any generated artifacts.

Keep it under a page. The point isn't storytelling — it's letting a human verify your work in five minutes and know what to look at next.

---

## Merged mode — parallel-subagent deep research (原 `deep-research`)

原顶层技能 `deep-research` 于 2026-09-10 并入本技能。它是一条**明确的 5 阶段流水线**：并行子 agent 做调研，**写作单点收口**（绝不让多个 agent 各写一段报告）。

### 何时用本模式

- 主题调研类任务、要"多方求证 + 带引用报告"
- **1-2 次搜索就能答的问题不要用**（直接 `web_search`）
- 枚举型任务（N 项 × M 字段，如"对比 20 个框架"）→ 用表格式分解，每个子 agent 负责一批

### 深度预算（硬上限）

| 模式 | 首轮子 agent | 最多追问轮次 | 来源目标 |
|---|---|---|---|
| quick | 2-3 | 0 | 8+ |
| standard（默认） | 3-5 | 1 | 15+ |
| deep | 5-8 | 2 | 25+ |

### 工作区（跨压缩持久）

所有状态落盘，不要只留在上下文里：

```text
research/<slug>/
├── brief.md         # 研究简报：所有阶段唯一对齐依据
├── findings/        # F1.md, F2.md ... 每个子 agent 一份结构化证据
└── REPORT.md        # 最终交付物
```

恢复时：重读 `brief.md` + 列出 `findings/`，跳过已完成角度继续。

### 五个阶段

1. **Scope** —— 最多一轮澄清（受众 / 时间范围 / 地区 / 要支撑的决策）；意图明确就直接把假设写进 `brief.md`，不要为问而问。
2. **Plan** —— 把简报拆成 3-8 个**互相独立**的角度（核心事实 / 近 12 个月进展 / 量化数据 / 反方与失败案例 / 一线实践 / 学术 / 关键玩家）；角度列表写入 `brief.md` 的 `## Angles`。
3. **Parallel** —— **同一条消息里**按角度并行起子 agent；提示词用 `references/deep-research/subagent-prompt.md` 的模板逐字复制，只替换 `{变量}`。每个子 agent 只研究一个角度，结构化写入 `findings/F<n>.md`（claim / quote / URL / date / confidence），**只回 3-5 行摘要**——原始页面内容不得进入主上下文。
4. **Reflect** —— 读遍 `findings/*.md`，对照 `brief.md` 找：哪些部分没有证据？哪些结论只靠单一来源？哪里来源互相冲突？有缺口且预算未尽 → 用更窄的 delta 查询再起一轮。
5. **Write（单点）** —— 只有你本人按 `references/deep-research/report.md` 一次成稿；每条非平凡结论带 `[n]` 引用且 URL 只来自 findings 文件（**不许凭记忆写**）；冲突来源并列呈现并标日期；单源结论标 `[single source]`，推测标 `[speculative]`；结尾给 Open questions 与编号 Sources（含访问日期）。deep 模式再补一遍怀疑式自审。

### 资源位置

- `references/deep-research/subagent-prompt.md` —— 子 agent 提示词模板（逐字复制）
- `references/deep-research/sources.md` —— 免 key 的免费数据端点清单
- `references/deep-research/report.md` —— 报告结构模板

> 说明：上游版本还提到一个外部脚本路径（`/Users/mi/...` 的 workflow 脚本），那是作者本机的文件，**本副本不引用**；需要无人值守运行就用 DSH 的 `workflow` 工具，脚本以 `script` 参数内联传入。