Manage AI Research Projects

Language: English | 简体中文

manage-ai-research-projects is an Agent Skill for organizing AI-era computational research projects. It helps Claude Code, Codex, and other skills-compatible agents create reproducible project structures, audit existing projects, separate AI-generated drafts from reviewed outputs, and keep scientific results traceable to data, code, parameters, environment, decisions, and manuscript deliverables.

The skill is based on a lightweight research project management template for computational biology, bioinformatics, AI-assisted scientific workflows, RAG/agentic screening projects, model benchmarks, and manuscript-oriented analysis projects.

Workflow

Repository Layout

.
└── manage-ai-research-projects/
    ├── SKILL.md
    ├── agents/openai.yaml
    ├── references/
    └── scripts/

The installable skill package is the manage-ai-research-projects/ directory.

Install

Claude Code

Install as a personal skill:

mkdir -p ~/.claude/skills
cp -R manage-ai-research-projects ~/.claude/skills/

Or install into one project:

mkdir -p .claude/skills
cp -R manage-ai-research-projects .claude/skills/

Invoke it directly in Claude Code with:

/manage-ai-research-projects

Codex

Install as a personal Codex skill:

mkdir -p ~/.codex/skills
cp -R manage-ai-research-projects ~/.codex/skills/

Then ask Codex to use $manage-ai-research-projects.

Use From GitHub

Users can download the repository ZIP or clone it, then copy the manage-ai-research-projects/ directory into their agent's skills directory.

Example:

git clone https://github.com/Devin-jun/Manage-AI-Research.git
cp -R Manage-AI-Research/manage-ai-research-projects ~/.claude/skills/

Capabilities

• Create a simplified research project skeleton with README.md, project.yaml, data/code/results/manuscript folders, and AI workflow records.
• Audit an existing project for missing metadata, weak reproducibility, unclear result traceability, ambiguous file names, and mixed AI draft/final outputs.
• Preserve raw data as immutable and separate it from processed analysis-ready files.
• Keep AI-generated code, prompts, agent traces, and reviewed outputs in distinct locations.
• Support manuscript and presentation delivery by keeping final assets separate from exploratory analysis results.

Validate Locally

Run the setup script on a temporary project:

python3 manage-ai-research-projects/scripts/init_research_project.py demo_project --root /tmp --title "Demo Project"

Audit the generated project:

python3 manage-ai-research-projects/scripts/audit_research_project.py /tmp/demo_project

Check Python syntax:

PYTHONPYCACHEPREFIX=/tmp/skill_pycache python3 -m py_compile \
  manage-ai-research-projects/scripts/init_research_project.py \
  manage-ai-research-projects/scripts/audit_research_project.py

License

MIT License. See LICENSE.