Set up knowledge-forge for this machine

Goal: give the user a one-command launcher for their OS, plus the right MCP/skill wiring for their AI tool.

1. Detect the platform

Determine the OS (Windows / macOS / Linux). You can ask the CLI for the exact instructions and target filenames:

knowledge-forge bootstrap --emit-instructions

2. Generate launchers

Fastest path (deterministic, no guessing):

knowledge-forge bootstrap --write

This writes, in the project root:

• Windows → kf.ps1 and kf.bat
• macOS / Linux → kf.sh (made executable)

Each launcher just cds into the project and forwards args to knowledge-forge, so the user can run ./kf.sh update / .\kf.ps1 update / kf.bat update.

If the user prefers hand-tuned scripts, write platform-idiomatic versions instead — but keep the same behavior (cd to project, forward args to knowledge-forge).

3. Wire up the AI tool

• Cursor: copy integrations/cursor/hooks.json + integrations/cursor/hooks/*

into the project's .cursor/, and merge integrations/cursor/mcp.json into .cursor/mcp.json. Copy skills/* into .cursor/skills/ (or .agents/skills/).

• Claude Code: copy integrations/claude-code/.mcp.json to the project root and

skills/* into .claude/skills/.

4. Verify

knowledge-forge --help

Then tell the user the exact one-line command to refresh their knowledge base on their OS (e.g. .\kf.ps1 update on Windows, ./kf.sh update on macOS/Linux).