luum-home/luum-cognitive-os/cognitive-os-init

Cognitive OS Init

META skill. Chains three atomic skills to fully initialize Cognitive OS for a project.

Overview

META skill that chains three atomic skills to fully initialize Cognitive OS for a project: detect-stack → generate-config → scaffold-project.

Atomic Skills

Each atomic skill can be invoked independently. Use this META skill when you want to run all three in sequence.

Execution Protocol

Step 1: Run /detect-stack

Invoke the detect-stack skill. It will:

• Scan go.mod, package.json, docker-compose.yml, pom.xml, etc.
• Detect languages, frameworks, infrastructure, services, and project type
• Write .cognitive-os/detected-stack.json

If it fails, stop and report the error. Do not continue to step 2.

Step 2: Run /generate-config

Invoke the generate-config skill. It will:

• Read .cognitive-os/detected-stack.json
• Generate or merge cognitive-os.yaml with detected infrastructure and quality gates

If it fails, stop and report the error. Do not continue to step 3.

Step 3: Run /scaffold-project

Invoke the scaffold-project skill. It will:

• Read .cognitive-os/detected-stack.json
• Create .claude/rules/architecture.md, constitutional-gates.md, services-config.md
• Create .claude/skills/sre-agent-config/SKILL.md, health-check/SKILL.md
• Create .claude/hooks/block-prod-urls.sh
• Update .cognitive-os/workflows/config/services.yaml

Step 4: Projection and primitive gate

After /scaffold-project, verify generated project hook settings resolve in the filtered install surface:

scripts/cos-install-projection-audit --json

If any generated project-local artifact is promoted into COS as a reusable primitive, run /primitive-authoring, add consumer/behavior evidence, and run primitive_scope_classifier.py --paths <promoted-primitive> --fail-contradictions --fail-low-confidence before committing it.

Step 5: Report

After all three skills complete, output the combined summary:

== Cognitive OS Init Complete ==
 
Phases completed:
  [x] detect-stack    → .cognitive-os/detected-stack.json
  [x] generate-config → cognitive-os.yaml
  [x] scaffold-project → .claude/ (7 files)
 
Next steps:
  1. Review generated files and adjust as needed
  2. Run /cognitive-os-test to verify configuration
  3. Run /sre-agent to verify service monitoring

Running Individual Phases

You can run any phase independently:

/detect-stack         # Only detect and record stack
/generate-config      # Only update cognitive-os.yaml (requires detected-stack.json)
/scaffold-project     # Only create .claude/ structure (requires detected-stack.json)

Important Notes

• This skill ONLY generates project-specific files in .claude/ — it does NOT modify Cognitive OS universal files
• Generated files are starting points — review and customize after running
• Re-running /cognitive-os-init will re-run all three phases (each will prompt before overwriting)
• The full pipeline is idempotent: running it twice produces the same result