Frascati Informatics Research Skill

A portable AI skill for evaluating and transforming Master's-level Informatics thesis ideas into research-oriented studies. It applies the OECD Frascati Manual 2015 to distinguish a knowledge contribution from a software implementation.

What It Does

• Evaluates novelty, creativity, uncertainty, systematic work, and transferability/reproducibility.
• Distinguishes basic research, applied research, and experimental development.
• Diagnoses implementation-centered proposals and transforms them into evidence-led research questions and experiments.
• Evaluates AI, ML, CV, NLP, LLM, Edge AI, datasets, benchmarks, prototypes, RAG, and software systems without assuming that any is automatically R&D.
• Applies the Chapter 2 exclusions (§2.8) to activities often mistaken for R&D: information services, routine testing and standardisation, specialised health care, policy-related studies, and programmatic evaluations.
• Keeps Frascati basis, S2 Informatics interpretation, and AI heuristic separate. The Manual does not discuss AI, ML, or LLMs explicitly; AI-specific reasoning is labelled as interpretation, not as a Frascati claim.

Repository Layout

skills/
└── frascati-informatics-research/
    ├── SKILL.md
    ├── references/
    │   ├── frascati-manual.md
    │   ├── software-rnd.md
    │   └── s2-research-framework.md
    └── examples/
        ├── implementation-to-research.md
        ├── cv-example.md
        ├── nlp-example.md
        ├── llm-example.md
        └── edge-ai-example.md

The repository deliberately does not redistribute the full OECD manual. references/frascati-manual.md is a limited, traceable locator ledger and is not a replacement for the official manual.

Locator pages use the PDF page numbers of the official OECD PDF (9789264239012-en.pdf, 402 pp.), matching the <!-- Page N --> markers in the local conversion. The printed folio at the foot of each page is the PDF page minus 2. Every locator in the ledger and examples is verified against the official PDF.

Install with npx skills

The skills CLI installs this skill directly from GitHub into any supported agent (CodeBuddy, Claude Code, OpenCode, Cursor, Codex, and many more):

# Install into the agents detected on your machine
npx skills add mhiqrambg/frascati-informatics-research

# Target specific agents
npx skills add mhiqrambg/frascati-informatics-research -a codebuddy -a claude-code -a opencode

# Global install, no prompts (e.g. for CI or a one-shot setup)
npx skills add mhiqrambg/frascati-informatics-research -g -a codebuddy -y

Preview the skill without installing, or use it once without installing:

npx skills add mhiqrambg/frascati-informatics-research --list
npx skills use mhiqrambg/frascati-informatics-research@frascati-informatics-research

Update or remove later:

npx skills update frascati-informatics-research
npx skills remove frascati-informatics-research

OpenCode

This repository includes opencode.json, which registers skills/ for project-local use. Restart OpenCode after cloning the repository.

To install globally:

mkdir -p ~/.config/opencode/skills
ln -s "$(pwd)/skills/frascati-informatics-research" \
  ~/.config/opencode/skills/frascati-informatics-research

Use cp -R rather than ln -s if a static copy is preferred. Restart OpenCode after installing.

Claude Code

Install the same self-contained folder project-locally:

mkdir -p .claude/skills
ln -s "$(pwd)/skills/frascati-informatics-research" \
  .claude/skills/frascati-informatics-research

Or install globally:

mkdir -p ~/.claude/skills
ln -s "$(pwd)/skills/frascati-informatics-research" \
  ~/.claude/skills/frascati-informatics-research

Restart Claude Code or start a new session after installing. The folder is deliberately self-contained so both tools use the same version.

Example Prompts

• Is my thesis idea of building a rice-disease diagnosis app with a CNN really R&D?
• Transform my idea of fine-tuning an LLM for education into a research question and an experimental design.
• Evaluate this Edge AI framework against the five Frascati criteria.
• Is this new dataset, model benchmark, and RAG prototype a knowledge contribution or an implementation?

Source and Attribution

The Frascati basis is OECD (2015), Frascati Manual 2015: Guidelines for Collecting and Reporting Data on Research and Experimental Development, OECD Publishing, Paris. DOI: 10.1787/9789264239012-en.

The full OECD manual is not included in this repository. For claims outside the verified ledger, users must consult an official OECD copy. The skill must report Source location not verified when that verification cannot be performed.

License and Sources

The repository's original content is licensed under MIT. OECD and other third-party material is not covered by that license. See SOURCES.md for source provenance, verification rules, and local-source handling.