AI4Math-Paper-Reading / paper-to-skill
AGENTS.md6 starsChanged 4 months agoMITView on GitHub
This repository implements a coding-agent-assisted paper-to-skill workflow for an AI4Math skill library. Codex is the reference operator for this repository, but the Skill layer should remain usable by other coding agents such as Claude Code, Gemini, OpenCode, Cursor, and similar agentic development tools. The goal is not to summarize papers, but to extract reusable mathematical research skills from papers. Prioritize the user-facing workflow Skill at skills/paper-to-skill-workflow. The default user input is paper.pdf. If the user provides paper.md, skip conversion and…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository7 other files it gives its agents.
- skills/paper-to-skill/CLAUDE.mdCLAUDE.md
- skills/paper-to-skill/.github/copilot-instructions.mdCopilot instructions
- ai4math-paper-readingSkill
- after-ocrSkill
- graph-theory-paper-readingSkill
- math-paper-readingSkill
- paper-to-skillSkill
The file
We split this AGENTS.md into the rules it gives an agent
9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Keep PDF conversion, extraction, retrieval, synthesis, and future automation as separate Skills rather than folding every responsibility …
- Very specific: line 23
Always preserve the original paper.pdf when provided and the full converted paper.md.
- Very specific: line 25
Every extracted skill must include source line references back to paper.md.
- Very specific: line 30
Keep platform-specific adapter files thin and point them back to skills/registry.yaml and the shared skills/ layer.
- Distinctive: line 31
Keep required YAML fields and templates explicit.
- Distinctive: line 32
Keep future Python automation compatible with the Skill workflow.
- Distinctive: line 33
Use direct mathematical checks and existing validation commands.
- Distinctive: line 34
Do not silently drop source text.
- Distinctive: line 35
Prefer small composable modules and deterministic output formats.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
Project Goal
This repository implements a coding-agent-assisted paper-to-skill workflow for an AI4Math skill library.
Codex is the reference operator for this repository, but the Skill layer should remain usable by other coding agents such as Claude Code, Gemini, OpenCode, Cursor, and similar agentic development tools.
The goal is not to summarize papers, but to extract reusable mathematical research skills from papers.
Current Scope
Prioritize the user-facing workflow Skill at skills/paper-to-skill-workflow.
The default user input is paper.pdf. If the user provides paper.md, skip conversion and extract directly.
The active coding agent is the extraction engine. The workflow should not require external LLM APIs.
Keep PDF conversion, extraction, retrieval, synthesis, and future automation as separate Skills rather than folding every responsibility into one file.
Source of Truth
Always preserve the original paper.pdf when provided and the full converted paper.md. Structured files are weak indexes, reports, or generated artifacts, not replacements for the original paper.
Every extracted skill must include source line references back to paper.md.
Engineering Rules
- Encode the extraction workflow as a portable Skill first.
- Keep platform-specific adapter files thin and point them back to
skills/registry.yamland the sharedskills/layer. - Keep required YAML fields and templates explicit.
- Keep future Python automation compatible with the Skill workflow.
- Use direct mathematical checks and existing validation commands; do not add a schema framework or test dependencies.
- Do not silently drop source text.
- Prefer small composable modules and deterministic output formats.
Done Means
A task is done only when:
- the portable Skill workflow is written or updated;
- a PDF paper can be processed end-to-end into Markdown, SkillCandidates, SkillCards, and a report;
- generated outputs match the expected schema;
- extracted skills include source evidence and review status.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

