unbrowse-ai Skill Bundle Fabric
Core job:
- turn a reusable workflow into one portable skill bundle that mines history, emits install/share/index artifacts, and writes executable host routing from one preset
Use this skill when:
- a workflow needs to become one portable installable bundle, not a pile of adjacent scripts
- the bundle should ship install commands, routing rules, share artifacts, registry metadata, and host memory together
- local chat history should help justify or refresh the bundled workflow surface
- the same bundle should work across Codex, Claude, or OpenClaw
- the bundle should be shareable by files now and indexable by backend later, with one entry skill over the child DAG
Do not use this skill for:
- building one domain skill with no bundle/share/index concern
- publishing runtime API routes to a marketplace
- writing free-form memory notes with no executable routing value
Workflow:
- Start from the bundle outcome: one entry skill, child skills, routes, share path, index metadata, host targets.
- Encode that once in
presets/<name>.json, including the bundle entry skill and dependency DAG. - Run
node scripts/fabricate-bundle.mjs --preset presets/<name>.json --out dist --host <host> --scope <scope>. - Inspect
history-report.json,bundle.json,share.json,registry-entry.json, and the host memory block. - Publish or share the generated artifacts with the target repo/host/backend.
Load-bearing rules:
- one bundle preset is the source of truth
- one entry skill owns the bundle-level job; child skills stay underneath it
- install commands, routing rules, share metadata, index metadata, and host memory stay derived from the same preset
- the memory block must say which skill to call, not just what to install
- dependency edges and alternatives belong in the preset, not hidden in prose
- support host-specific root files:
AGENTS.md,CLAUDE.md,MEMORY.md - build artifacts must be backend-ready even if the backend is not live yet