Research-Skills
AGENTS.md9 starsChanged 6 months agoMITView on GitHub
This repo is a skills collection for OpenCode/Hermes — not a code project with tests or builds. The skills live under skills/ and are loaded on demand via the skill tool. Read skills/for_system_prompt.md — it's the meta-system prompt for the Academic Research Copilot role. Inject its content into your system prompt before using paper-writing or patent-writing. Two umbrella skills cover the full academic workflow: Both reference sub-modules in skills/paper-writing/references/. MIT License © 2026 nors1993
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 31,700 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 repository8 other files it gives its agents.
- arxivSkill
- blogwatcherSkill
- docxSkill
- fireworks-tech-graphSkill
- opencode-integrationSkill
- paper-writingSkill
- patent-writingSkill
- markitdown-converterSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
Always ask user for "current time point" before writing
- Distinctive: line 35
If keywords include "bibliometric", "文献计量", "VOSviewer", etc., do NOT run the standard 7-step workflow.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
Research-Skills AGENTS.md
This repo is a skills collection for OpenCode/Hermes — not a code project with tests or builds. The skills live under skills/ and are loaded on demand via the skill tool.
First Things
Read skills/for_system_prompt.md — it's the meta-system prompt for the Academic Research Copilot role. Inject its content into your system prompt before using paper-writing or patent-writing.
Primary Entry Points
Two umbrella skills cover the full academic workflow:
skills/paper-writing/SKILL.md— 7-step paper writing pipeline (intent→feasibility→deep research→drafting→consistency→plagiarism→humanize)skills/patent-writing/SKILL.md— 7-step patent drafting pipeline (same structure)
Both reference sub-modules in skills/paper-writing/references/.
Important Gotchas
DO use paper-writing SKILL.md directly, NOT research-* sub-skills
The README documents research-intent, research-idea-parser, etc. as separate directories — they don't exist as separate dirs anymore. The full pipeline logic is embedded directly in paper-writing/SKILL.md and patent-writing/SKILL.md. Route via for_system_prompt.md instead.
docx generation pitfalls (from paper-writing SKILL.md)
- For long Chinese documents (>20,000 characters), use python-docx not docx-js
- docx-js has f-string escaping issues with Chinese Unicode (
\u201c/\u201dconflict with JS string delimiters) - Subagent timeout: do NOT delegate final docx generation to subagents — they will time out (600s+). Generate in the main session.
Data currency (paper-writing SKILL.md)
- Distinguish Current Status (past real data) from Trends/Forecast (future-only)
- Always ask user for "current time point" before writing
- Label data sources with cutoff dates
- SIA annual report, FAA, planet4589.org, UCS satellite database are top data sources
Bibliometric papers
If keywords include "bibliometric", "文献计量", "VOSviewer", etc., do NOT run the standard 7-step workflow. Read references/bibliometric-paper-workflow.md first — the flow is fundamentally different (real data collection via CrossRef API, analysis + visualization, data export as CSV).
arXiv skill
Uses curl + arXiv REST API (no API key). Parse Atom XML output with inline Python.
Language
The repo is bilingual (Chinese/English). Primary documentation is in Chinese. Questions may come in either language.
Available Skills Under skills/
| Directory | Description |
arxiv/ | arXiv paper search (curl + REST API) |
blogwatcher/ | RSS/Atom feed monitoring |
docx/ | Word document creation/editing |
paper-writing/ | Umbrella: academic paper writing (main) |
patent-writing/ | Umbrella: patent drafting |
opencode-integration/ | Use OpenCode as subagent via ACP |
fireworks-tech-graph/ | Technical diagrams as SVG+PNG |
software-development/markitdown-converter/ | Doc→Markdown conversion |
| (18 DESCRIPTION.md-only) | Categorized domains: research, apple, creative, data-science, diagramming, domain, email, gaming, gifs, github, inference-sh, mcp, media, mlops, note-taking, productivity, smart-home, social-media |
Tools Required
curl— arXiv APIblogwatcher-cli— RSS monitoringpython-docx/docx-js— .docx generationpandoc— doc conversionmatplotlib— chart generation (install via pip with--break-system-packages)cairosvg— SVG→PNG conversion (fireworks-tech-graph)markitdown— doc→markdown conversion (installed at~/.local/bin/markitdown)
Licensing
MIT License © 2026 nors1993
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.