llms.txt

Machine-readable framework index for AI assistants — following the llms.txt convention.

📄 What is llms.txt?

llms.txt is an emerging convention that provides a machine-readable index of a project's documentation. It lets Large Language Models (and any tool that works with them) instantly understand a project's purpose, structure, and capabilities — without crawling dozens of files.

OMNISKILL generates two files:

File Purpose Size
llms.txt Concise index — skill names, agent roles, pipeline chains, bundle contents ~10–15 KB
llms-full.txt Complete dump — every SKILL.md, AGENT.md, SYNAPSE.md, pipeline YAML, bundle YAML, and docs page ~500 KB – 3 MB

🧩 What OMNISKILL Generates

Concise Index (llms.txt)

Full Dump (llms-full.txt)

Contains the complete text content of every component file: README.md, omniskill.yaml, every SKILL.md, AGENT.md, SYNAPSE.md, pipeline YAML, bundle.yaml, and documentation page.

⚡ How to Generate

CLI Command (recommended)

# Generate both files (default)
omniskill generate llms-txt

# Generate only the concise index
omniskill generate llms-txt --concise

# Generate only the full dump
omniskill generate llms-txt --full

# Write to a custom directory
omniskill generate llms-txt --output ./dist/

# JSON output (for scripting)
omniskill --json generate llms-txt

# Verbose output
omniskill --verbose generate llms-txt

Standalone Script

Works without the omniskill package installed — only requires Python 3.10+ and PyYAML:

# Generate both files
python scripts/generate-llms-txt.py

# Concise only
python scripts/generate-llms-txt.py --concise

# Full only
python scripts/generate-llms-txt.py --full

# Custom output directory
python scripts/generate-llms-txt.py --output ./dist/

🚀 Deployment

When you push to the master branch, GitHub Actions automatically generates and deploys both files. After deployment:

✅ Validation

Check whether your local files are up to date:

# Via CLI
omniskill validate --check-llms-txt

# Via standalone script
python scripts/validate.py --check-llms-txt

The staleness check is warning-only — it never causes validation failures.