Skip to content

Light Skills v1.0.0

Latest
Latest

Choose a tag to compare

View all tags
@Light0305 Light0305 released this 06 Jul 03:00
· 1 commit to master since this release
v1.0.0
e8371f8

Light Skills v1.0.0 Release Notes

Release date: 2026-07-05

Light Skills v1.0.0 is the first public release of the rebuilt Light Skills package: a domain-independent AI workflow skill library for research, competitions, and innovation projects.

The goal of this release is to move AI-assisted research from one-off chat prompts toward inspectable, reproducible, and deliverable workflows.

Highlights

  • 23 public skills for the research and innovation workflow.
  • Cross-client support for Codex, Claude Code, and OpenCode.
  • Installation helper for supported agent-skill layouts.
  • Shared evidence, consistency, review, and workflow contracts.
  • Programmatic scientific figure expectations with Python and R.
  • LaTeX-oriented paper typesetting workflow.
  • End-to-end research demo and reproducible paper demo.
  • MIT license.

Who is this for?

Light Skills is designed for users who want to use AI agents in real project work, including:

  • students doing research training or course projects;
  • researchers drafting experiments, figures, and papers;
  • competition teams building reproducible project deliverables;
  • developers building research demos or technical prototypes;
  • users preparing patent-disclosure or software-copyright materials.

What changed from the earlier Light project?

The earlier Light work emphasized broad research-process coverage. The v1.0.0 public package reorganizes that work into a cleaner, public, reusable skill library.

Major changes:

  • consolidated overlapping skills into stronger user-facing workflows;
  • removed private-database framing from the public package;
  • clarified that the public version does not require private MCP servers or local knowledge bases;
  • added explicit support for Codex, Claude Code, and OpenCode layouts;
  • added reproducible demo artifacts for papers and figures;
  • strengthened honesty boundaries around citations, unknown facts, figures, and innovation claims.

Skill catalog

The release includes skills for:

  • citation integrity;
  • consistency checks;
  • data engineering;
  • experiment coding;
  • scientific figures;
  • file reading;
  • frontend design;
  • idea critique;
  • idea generation;
  • literature search;
  • project memory and project management;
  • orchestration;
  • paper writing;
  • patent disclosure;
  • project structure;
  • research ethics;
  • research planning;
  • result analysis;
  • review and rebuttal;
  • software copyright;
  • system design;
  • LaTeX/typesetting;
  • venue matching.

Installation

Clone the repository:

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

Install into a supported project layout:

python scripts\bootstrap_agent_skills.py --targets agents --mode auto --force
python scripts\bootstrap_agent_skills.py --targets claude --mode auto --force
python scripts\bootstrap_agent_skills.py --targets opencode --mode auto --force

Check installation:

python scripts\bootstrap_agent_skills.py --check-only

Optional environment layers

Base usage expects:

  • Git
  • Python 3.10+

Some workflows benefit from optional tools:

  • LaTeX for paper compilation and template checks;
  • R for scientific visualization workflows;
  • Python plotting/data packages for reproducible figures and analysis.

If an optional tool is unavailable, skills should report it honestly rather than pretending the workflow succeeded.

Integrity boundaries

Light Skills v1.0.0 intentionally avoids several common AI-tool overclaims.

It does not promise:

  • guaranteed paper acceptance;
  • guaranteed patents or software-copyright registration;
  • automatic true scientific innovation;
  • verified facts without retrieval or inspection;
  • authentic DOI values, links, or citations without checking;
  • scientific figures generated by image-only AI tools.

It does aim to help users:

  • search more systematically;
  • critique ideas earlier;
  • separate target, background, method, and evidence;
  • build reproducible figures;
  • organize paper drafts and LaTeX outputs;
  • check citations and unknown claims;
  • leave reviewable traces for human judgment.

Known limitations

  • The public package is still early and needs more community feedback.
  • The repository currently emphasizes Windows-first commands.
  • Some workflows require the user's local environment to provide LaTeX, R, Python packages, or agent-client support.
  • Skills guide workflows; they do not replace domain experts, supervisors, reviewers, lawyers, or ethics committees.

Maintainer notes

This release prepares the project for public use, issue reports, and future community contributions. The next maintenance focus is to improve tests, examples, release gates, and compatibility checks while keeping claims aligned with what the repository can actually support.

Assets 2
Loading