knowledge-system-bootstrap

Knowledge System Bootstrap

Use this when the user wants a project to stop relying on chat memory and start behaving like a sane system.

What this skill does

Scaffolds a complete wiki-first knowledge system (30 files):

• docs/wiki/ — 8 wiki pages with YAML frontmatter (title, source, created, tags, status, optional provenance fields)
• manifests/raw_sources.csv — raw file index (never raw files themselves)
• scripts/ — 11 validation and utility scripts:
  • wiki_check.py — structure + broken links + frontmatter enforcement
  • ingest_raw.py — scan a local raw root, dedupe, update manifest, and build a low-cost intake report with table-level diff summaries
  • raw_manifest_check.py — manifest integrity
  • untracked_raw_check.py — finds orphan PDFs/Excel/images not in manifest
  • provenance_check.py — content hash freshness (source_hash in frontmatter) with strict unresolved-source failures
  • stale_report.py — report wiki pages that are stale, missing hashes, unresolved, or blocked by manifest status
  • delta_compile.py — generate manual draft stubs for stale/new raw instead of auto-overwriting wiki content
  • version_check.py — auto-checks GitHub for new LLM-wiki releases
  • upgrade.sh — one-command upgrade (scripts only, never touches wiki content)
  • init_raw_root.py — create local raw directory structure
  • export_memory_repo.py — export wiki to separate memory repo
• Platform configs: AGENTS.md, CLAUDE.md, .cursorrules, .windsurfrules
• .claude/commands/ — Claude Code slash commands: /wiki-check, /wiki-upgrade, /wiki-status
• .github/workflows/wiki-lint.yml — CI smoke test

Default stance

• compile-first
• writeback is mandatory
• medium-sized projects use wiki before heavy RAG
• Obsidian is optional, markdown is not
• Idea / Intent outranks Code
• full audit = disaster recovery, not normal ops

When to use it

• set up a durable project memory layer
• stop losing context between sessions
• separate GitHub knowledge from raw binary junk
• create a repeatable wiki/raw/manifests structure for a new repo
• migrate an existing repo toward this model

Do not use it for tiny throwaway demos.

Workflow

1. Identify the target repo root and project name.
2. Preview: python3 scripts/bootstrap_knowledge_system.py /path/to/repo "Name" --dry-run
3. Run: python3 scripts/bootstrap_knowledge_system.py /path/to/repo "Project Name"
4. Initialize local raw root: python3 scripts/init_raw_root.py
5. Intake raw if needed: python3 scripts/ingest_raw.py
6. Validate: python3 scripts/wiki_check.py && python3 scripts/raw_manifest_check.py && python3 scripts/stale_report.py
7. If stale/new raw exists, scaffold manual recompilation drafts: python3 scripts/delta_compile.py --write-drafts

Existing repo migration

If the repo already has docs or a CLAUDE.md:

• Run bootstrap with default settings (skips existing files)
• Move existing docs into docs/wiki/ manually
• Register raw files in manifests/raw_sources.csv
• Merge customized config rules with generated templates

Upgrade

Existing bootstrapped projects can upgrade in place:

bash scripts/upgrade.sh

If the project predates scripts/upgrade.sh, use the public repo wrapper once:

git clone https://github.com/Ss1024sS/LLM-wiki.git
cd LLM-wiki
bash scripts/upgrade.sh /path/to/your-project

Updates validation scripts and CI only. Wiki content and customized configs are never touched.

Bundled resources

• Script: scripts/bootstrap_knowledge_system.py
• Reference: references/playbook.md (points to docs/knowledge-system-playbook.md)