---

What can it help you do?

Light Skills is a public, domain-agnostic skill pack for turning a research or innovation project from a vague idea into auditable deliverables.

Your situation	What Light Skills helps with
I only have a topic	Ask clarifying questions, define constraints, and break the project into stages
I have an idea but do not know whether it is new	Search related work, split target/background, find collisions and reviewer attacks
I need experiments or data analysis	Design data flow, experiment matrix, scripts, tests, result analysis, and robustness checks
I need an English paper	Structure the story, figures, citations, LaTeX typesetting, and pre-submission checks
I need scientific figures	Generate Python/R figures programmatically with size, typography, and honesty checks
I need a demo or competition interface	Design frontend demos, system structure, interactions, and presentation material
I need patent or software copyright material	Draft disclosure materials, technical schemes, embodiments, and document lists
I need to continue across chats	Keep a project ledger of goals, decisions, artifacts, unverified claims, and next steps

Why it fits research projects

• Read before writing: inspect files, data, logs, and paper sources before making claims.
• Mark unknowns honestly: facts, DOI, links, venue rules, and software versions are checked instead of guessed.
• Generate reproducible figures: paper, data, and experiment figures are produced with Python/R code.
• Ask at decision points: topic choice, novelty, evidence strength, venue target, and further investment need user confirmation.
• No private knowledge base required: the public package does not require MCP or a local database; current facts are checked at task time.

Install

Enter the repository:

git clone https://github.com/Light0305/Light-skills.git
cd Light-skills
$env:PYTHONUTF8="1"

Codex

# Project-level: $REPO\.agents\skills
$env:PYTHONUTF8="1"
python scripts\bootstrap_agent_skills.py --targets agents --mode auto --force

# Global: $HOME\.agents\skills
New-Item -ItemType Directory -Force "$HOME\.agents\skills" | Out-Null
Copy-Item -Recurse -Force .\skills\* "$HOME\.agents\skills\"

Claude Code

# Project-level: $REPO\.claude\skills\<skill>\SKILL.md
$env:PYTHONUTF8="1"
python scripts\bootstrap_agent_skills.py --targets claude --mode auto --force

# Global: $HOME\.claude\skills
New-Item -ItemType Directory -Force "$HOME\.claude\skills" | Out-Null
Copy-Item -Recurse -Force .\skills\* "$HOME\.claude\skills\"

OpenCode

# Project-level: $REPO\.opencode\skills\<skill>\SKILL.md
$env:PYTHONUTF8="1"
python scripts\bootstrap_agent_skills.py --targets opencode --mode auto --force

# Global: $HOME\.config\opencode\skills
New-Item -ItemType Directory -Force "$HOME\.config\opencode\skills" | Out-Null
Copy-Item -Recurse -Force .\skills\* "$HOME\.config\opencode\skills\"

Check the installation:

$env:PYTHONUTF8="1"
python scripts\bootstrap_agent_skills.py --check-only

Environment

Base

• Git
• Python 3.10+
• On Windows, set $env:PYTHONUTF8="1" before running Python scripts.

LaTeX

winget install --id MiKTeX.MiKTeX --accept-package-agreements --accept-source-agreements
latexmk -v
pdflatex --version
xelatex --version
biber --version

Use this for paper typesetting, PDF compilation, and template checks. light-typesetting supports latexmk, pdfLaTeX, XeLaTeX, LuaLaTeX, BibTeX, and Biber. If a tool is missing, it reports UNAVAILABLE instead of pretending that a PDF was delivered.

R

winget install --id RProject.R --accept-package-agreements --accept-source-agreements
Rscript -e "install.packages(c('ggplot2','scales'), repos='https://cloud.r-project.org')"
$env:PYTHONUTF8="1"
python skills\light-figure\scripts\r_ggplot.py --detect

Use this for ggplot2 scientific figures. If R is missing, the figure skill should ask whether to continue with a Python fallback or install/configure R.

Start from your current state

Current state	Suggested entry
Topic only	/light-orchestrator I want to turn this direction into a submittable English paper. Ask necessary questions first, then propose stages, deliverables, risks, and decision gates.
Existing idea	$light-idea-critique Critique this idea: novelty, falsifiability, related work, strongest counterexamples, reviewer risks, and validation plan.
Existing files	$light-file-reading Read this project directory first. List key files, completed work, unverified claims, risks, and next steps.
Literature search	$light-literature-search Build a search strategy, keyword expansion, evidence map, and related-work boundary for this question.
Experiment design	$light-research-plan Give an experiment matrix, data needs, metrics, failure conditions, and minimum viable validation.
Figures	$light-figure Plan paper figures from these data. Require programmatic generation, reproducibility, colorblind-safe palettes, and clear labels.
Paper writing	$light-paper-writing Organize an English manuscript from existing evidence, with contributions, limitations, and self-review checks.
Typesetting	$light-typesetting Build and preflight the current LaTeX source, figures, BibTeX, and venue template.
Interface/demo	$light-frontend-design Design a demo page, component structure, interactions, and presentation focus for this research project.

Skill map

Module	Skills
Orchestration and continuity	light-orchestrator, light-memory-pm, light-file-reading, light-project-structure
Ideas and literature	light-literature-search, light-idea-generation, light-idea-critique, light-research-plan
Data and experiments	light-data-engineering, light-experiment-coding, light-result-analysis
Paper delivery	light-paper-writing, light-citation, light-consistency, light-typesetting, light-venue-matching, light-review-rebuttal
Figures and demos	light-figure, light-frontend-design, light-system-design
Integrity and IP	light-research-ethics, light-patent-disclosure, light-software-copyright

Complete skill catalog

Skill	Main use	Typical outputs
light-orchestrator	Main controller: understand the task, ask necessary questions, select the skill chain, and set user decision gates	Stage plan, skill routing, decision checkpoints, workflow ledger
light-memory-pm	Project ledger and cross-session continuation without storing private memory in the public repository	Project card, handoff card, decision log, resume prompt
light-file-reading	Read papers, PDFs, Word files, slides, spreadsheets, images, and project files	File inventory, understanding notes, extraction-quality report, unverified-claim list
light-project-structure	Scaffold and govern research/software project directories for maintainability and reproducibility	Project scaffold, directory policy, governance rules, structure check
light-literature-search	Build search strategies, expand keywords, track evidence boundaries, and map related work	Search strings, evidence map, literature table, PRISMA-style flow record
light-idea-generation	Generate candidate research ideas from literature gaps, cross-domain analogies, and constraints	Idea cards, gap evidence, genealogy analysis, candidate ranking
light-idea-critique	Critique novelty, falsifiability, feasibility, and fatal flaws before committing to an idea	Go/no-go verdict, counterexample list, revision roadmap, novelty-evidence gate
light-research-plan	Convert a question into an executable study plan with hypotheses, variables, controls, and failure trees	Experiment matrix, preregistration draft, sample-size/power checks, reproducibility plan
light-research-ethics	Check ethics, authority, consent, data boundaries, and research-integrity risks	Ethics risk table, authority lifecycle check, retraction/overlap/anomalous-text alerts
light-data-engineering	Assess data identity, access, quality, splits, leakage, feasibility, and drift risks	Data card, quality gate, leakage check, feasibility report
light-experiment-coding	Build reproducible experiment code, configuration, tests, and run records	Experiment scaffold, config schema, seed audit, run manifest
light-result-analysis	Run statistical analysis, method-compatibility checks, leakage/overfit checks, and result interpretation	Analysis report, statistical tests, method-compatibility check, result card
light-figure	Plan and generate paper/data figures programmatically, with Python and R support	Figure plan card, Python/R figures, export package, visual-honesty check
light-paper-writing	Draft manuscript structure, argument chains, contributions, limitations, and self-review from existing evidence	IMRaD/conference draft, claim-evidence binding, self-review checklist, polished manuscript
light-citation	Verify citation truth, DOI, links, locators, and claim-citation binding	Citation registry, four-gate verification, suspicious-reference list, repair suggestions
light-consistency	Check consistency across the paper, figures, slides, code, and supplementary materials	Glossary, fact bindings, metric/method locks, cross-material consistency report
light-typesetting	Typeset with LaTeX templates, compile PDFs, inspect logs, and preflight submissions	Compilable LaTeX/PDF, template adaptation, build log, submission-readiness check
light-venue-matching	Match journals/conferences using topic fit, evidence, risk, and privacy constraints	Venue candidate table, fit ranking, risk notes, user-selection record
light-review-rebuttal	Decompose reviews, plan extra experiments, manage commitments, and draft responses	Response matrix, commitment ledger, experiment-request gate, response letter
light-frontend-design	Design interfaces, components, and presentation experiences for research projects, competitions, or software demos	Page structure, component plan, motion suggestions, accessibility/browser QA
light-system-design	Design software architecture, APIs, data models, migrations, and readiness checks	Architecture package, OpenAPI/schema, migration strategy, design-readiness report
light-patent-disclosure	Organize invention points, prior-art differences, and patent disclosure materials without replacing legal advice	Patent interview notes, search leads, disclosure evidence packet
light-software-copyright	Prepare software-copyright materials, source-deposit planning, and completeness checks	Software-copyright packet, source-deposit plan, materials completeness check

Research workflow

Light Skills is not designed to let an agent blindly run the whole project end to end. It breaks research into auditable stages with explicit checkpoints, rollback paths, and user decisions.

Stage	What Light mainly does	Common skills
Intake and understanding	Read papers, tables, images, and project files; define task boundaries and maintain a project ledger	file-reading, memory-pm, project-structure
Novelty and ideas	Search literature, generate candidate ideas, and critique novelty, feasibility, and fatal flaws	literature-search, idea-generation, idea-critique
Research design	Specify hypotheses, variables, controls, sample size, failure trees, and reproducible study plans	research-plan, research-ethics
Data and experiments	Prepare data, check leakage and quality, implement experiments, and analyze results	data-engineering, experiment-coding, result-analysis
Paper delivery	Generate reproducible figures, write the manuscript, verify citations, check consistency, and typeset LaTeX	figure, paper-writing, citation, consistency, typesetting
Submission and IP	Match venues, prepare rebuttals, and organize patent or software-copyright materials	venue-matching, review-rebuttal, patent-disclosure, software-copyright
Demo and software	Add frontend demos and system design when a project, competition, or product needs an interface	frontend-design, system-design

Paper demo

A paper demo example in environmental chemistry / photocatalytic kinetics, showing the complete deliverable shape from synthetic data, analysis, programmatic figures, and LaTeX PDF.

_{Read PDF}

Figure gallery

The gallery below is generated with Python and R.

Feedback and support

• Email: [email protected]
• GitHub: @Light0305
• Issues, PRs, and usage feedback are welcome.

WeChat donation	WeChat official account

License

This project is released under the MIT License.