omnigraph-intel-bootstrap

omnigraph-intel-bootstrap is a packaged skill that sets up a new SPIKE-based graph end-to-end. It handles the decision of demo-vs-custom, elicits domain context from the user, adapts the schema and enums, researches initial seed content from the live web, and executes init + load.

When to use it

Trigger phrases the skill is tuned for: set up Omnigraph, bootstrap a new graph, create a new SPIKE starter, I want to track X industry, initialize intel for Y, new graph for Z domain, start a new context graph.

What it does

After a Step 0 pre-flight check (port :8080 collision, PATH for the installed binaries), the skill asks Step 1, demo or custom. And then runs one of two paths:

Demo (Path A). Clone the cookbook, init, load the reference AI-industry seed (5 patterns, 15 signals, 111 nodes / 148 edges). Ready to query in roughly 30 seconds.
Custom (Path B). Seven-phase flow for a new domain:
1. Domain identification
2. Key questions the graph should answer
3. Schema and enum adaptation
4. Initial web research for seed content
5. Seed assembly and validation
6. init + load + verification reads
7. Hand-off to the omnigraph-best-practices skill

The custom path enforces one rule that does most of the work: do not fabricate signals or dates. Every seeded signal must come from a real, linkable source.

Prerequisites

Docker (install). RustFS runs in a container.
RustFS on 127.0.0.1:9000. The local-rustfs-bootstrap.sh script in the main repo sets this up:
```
curl -fsSL https://raw.githubusercontent.com/ModernRelay/omnigraph/main/scripts/local-rustfs-bootstrap.sh | bash
```
The bootstrap installs binaries under <workdir>/.omnigraph-rustfs-demo/bin/ . Not on PATH by default. Add it or invoke binaries by absolute path.
Port :8080 free for the cookbook server (the bootstrap auto-starts its own server there. Stop it or use --bind 127.0.0.1:8090 for yours).
The omnigraph-cookbooks repo cloned somewhere on disk. The custom path copies industry-intel/ as a template; the demo path runs from industry-intel/ directly.
omnigraph CLI >= 0.3.1.

AWS env vars

init and load talk to S3-backed storage directly. Both require the seven AWS variables pointed at RustFS. Kept in .env.omni (git-ignored):

AWS_ACCESS_KEY_ID=rustfsadmin
AWS_SECRET_ACCESS_KEY=rustfsadmin
AWS_REGION=us-east-1
AWS_ENDPOINT_URL=http://127.0.0.1:9000
AWS_ENDPOINT_URL_S3=http://127.0.0.1:9000
AWS_ALLOW_HTTP=true
AWS_S3_FORCE_PATH_STYLE=true

AWS_ALLOW_HTTP and AWS_S3_FORCE_PATH_STYLE are mandatory. Omitting either produces a builder error from lance-io at init/load time.

If this is a fresh RustFS instance, create the bucket once:

aws --endpoint-url http://127.0.0.1:9000 s3 mb s3://omnigraph-local

Deep-dive references

The skill ships five self-contained reference files:

Reference	Purpose
`demo-setup.md`	Demo-path runbook with bucket creation, init/load, expected output, verification queries
`custom-domain.md`	Custom-path elicitation flow. What to ask the user and how to shape the schema from the answers
`schema-adaptation.md`	Editing the cookbook schema and enums for a new domain
`domain-examples.md`	Worked examples. Biotech, fintech, geopolitics, climate, more
`research.md`	Phase 4 web research playbook + the "do not fabricate" rule

Installing

npx skills add ModernRelay/omnigraph-cookbooks@omnigraph-intel-bootstrap

After installation, invoke by asking the agent to "set up an Omnigraph SPIKE graph" (or any of the trigger phrases above). When bootstrap completes, switch to omnigraph-best-practices for day-to-day operations.

Reference

Full SKILL.md: skills/omnigraph-intel-bootstrap/SKILL.md
SPIKE framework and schema: Industry Intelligence guide