Research Like A Clown — Development skill for Claude Code
A Claude Code writing skill with 28 documented research error patterns and traceable source records.
How to install Research Like A Clown
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ShipOfShame/Research-Like-a-Clown and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Research Like A Clown does
A Claude Code writing skill with 28 documented research error patterns and traceable source records.
Alternatives in Development
- Scientific Agent Skills — A set of ready to use Agent Skills for research, science, engineering, analysis, finance and writing 18.7k ★
- Codex Claude Academic Skills — 本仓库包含三个面向学术科研人员的Skills，覆盖从文献阅读、论文写作到科学计算的完整研究工作流 3.2k ★
- Claude Scholar — Semi-automated research assistant for academic research and software development 3k ★
README
Research Like a Clown
[](https://code.claude.com/docs/en/skills) [](references/patterns.md) [](references/evidence.json) [](https://github.com/ShipOfShame/Research-Like-a-Clown/stargazers)
**Paper-derived error patterns as a writing skill, designed first for Claude Code.**
[Quick start](#install-in-claude-code) · [Pattern catalog](references/patterns.md) · [Source records](references/evidence.json) · [Contributing](CONTRIBUTING.md) · [简体中文](README.zh-CN.md)
The initial case material comes from review records of papers coauthored by **Ironieser (Sixun Dong)**. Research Like a Clown distills the recorded paper and code issues into a writing skill that deliberately generates passages containing those error mechanisms. Original references and their scope are retained for checking each pattern.
What is included
| Area | Representative patterns |
|---|---|
| Results and metrics | Invalid aggregates, switched metric meanings, scores assigned to the wrong condition |
| Mathematical definitions | Identity state transitions, incompatible function domains, incorrect denominators |
| Evaluation | Split overlap, test-guided selection, normalization that removes errors |
| Implementation and artifacts | Missing evaluator inputs, code/specification mismatches, unexecuted revisions |
| Reporting and availability | Inconsistent model identifiers, vocabulary/sequence confusion, incomplete materials |
The **2026-09-26 review snapshot** contains **33 issue records across 19 papers**, distilled here into **28 patterns**. Repeated mechanisms are grouped; one record containing two different defects maps to two patterns. This is coverage
Related Skills
Context Graph
Evidence-first Knowledge Engineering Skill for Codex and Claude Code—turning sources into traceable records, v
Discovery Synthesise
Turn raw research into coded, traceable findings with confidence levels and a disconfirming-evidence section.
Contextualize
Analyze source document(s) to understand their purpose, research the project deeply for context and patterns,
Patent Skills
Open-source Agent Skills for patent and innovation research. Discover distinctive technical patterns in a code
Claude Code Info
Leaked Claude Code source tree, documented: every file under claude/src with generated summaries, exports, dep
Tapeback
Like rewinding a tape 🎞️ — tapeback automatically records every Claude Code agent action that changes your co
Related Agents
Prose Humanizer
Rewrites prose to remove AI writing patterns. Use when editing documentation, READMEs, PR descriptions, commit
Source Scout
Decides which archives, indexes and databases could plausibly hold records for a given person or family. Reads
Feynman
Audit investment-research facts, figures, dates, periods, and quotes against traceable primary sources. Use be
