Multi-skill repository for GeoVerse AI agent capabilities.
| Name | Description | Status |
|---|---|---|
| generate-region-layout | Facility grid / rectangle → deterministic layout planning | ✅ Production |
# Validate a specific skill
node shared/check-skill-package.mjs generate-region-layout
# Validate all skills (same as CI)
for dir in skills/*/; do node shared/check-skill-package.mjs "$(basename "$dir")"; doneInstall only skills/generate-region-layout into $CODEX_HOME/skills/generate-region-layout; do not leave a second backup folder with the same Skill name under $CODEX_HOME/skills. Configure the independently published MCP server with a pinned version:
codex mcp add geoverse-layout -- npx -y @geoverse/layout-mcp-app@0.5.0From the installed Skill directory, run node scripts/check-agent-installation.mjs to reject missing, duplicate, or stale copies. Then run node scripts/check-runtime.mjs and confirm Engine 0.7.x, MCP App 0.5.x, and ui://geoverse/region-plan-map-v2.html. Start a new Agent session after installation or MCP changes.
cp -r templates/ skills/<new-skill-name>/
# Edit SKILL.md, agents/openai.yaml, evals/evals.json
# Then validate:
node shared/check-skill-package.mjs <new-skill-name>See templates/README.md for detailed instructions.
geoverse-skills/
├── README.md ← You are here
├── CONTRIBUTING.md ← How to add or modify skills
├── shared/ ← Cross-skill utilities and schemas
│ ├── check-skill-package.mjs
│ ├── skill-schema.json
│ └── eval-schema.json
├── templates/ ← Scaffold for new skills
│ ├── SKILL.md
│ ├── agents/openai.yaml
│ ├── evals/evals.json
│ ├── references/
│ └── scripts/
├── skills/ ← All skills live here
│ └── <skill-name>/
│ ├── SKILL.md ← Definition and workflow
│ ├── agents/ ← Agent interface config
│ ├── evals/ ← Evaluation cases
│ ├── references/ ← Domain knowledge
│ └── scripts/ ← Executable tooling
├── docs/ ← Cross-cutting documentation
└── .github/workflows/ci.yml ← Unified CI
Every skill under skills/ must follow this structure:
| Path | Required | Purpose |
|---|---|---|
SKILL.md |
✅ | Frontmatter (name, description) + workflow documentation |
agents/openai.yaml |
✅ | Agent display name, short description, default prompt |
evals/evals.json |
✅ | Evaluation cases with prompts and expected outputs |
references/ |
✅ | Domain-specific contracts, capabilities, examples |
scripts/ |
Recommended | Executable tooling (validators, generators, adapters) |
The unified CI workflow automatically discovers all skills under skills/ and runs:
- Structure validation via
shared/check-skill-package.mjs - Syntax check on all
.mjsscripts - Domain-specific validation if sample data and validators are present
- Runtime-resolution tests and a clean npm-install smoke test for versioned GeoVerse packages
This project is licensed under the Business Source License 1.1. See the LICENSE file for details.