| name | obsidian-intelligence-system |
| description | Scaffold, audit, and evolve an Obsidian vault into a reusable intelligence system or product second brain. Use when a user wants to turn project documents, repos, links, notes, PDFs, or recurring entity records into a linked vault with YAML frontmatter, internal links, dashboards, templates, Bases, Canvas, hub nodes, intelligence summaries, and research-to-action synthesis. |
Obsidian Intelligence System
Overview
Use this skill when the goal is not just to store notes in Obsidian, but to turn a vault into a working memory system for a project, team, product, research corpus, or operational portfolio.
When the user is trying to scale this beyond one product, pair this skill with:
$CODEX_HOME/skills/product-intelligence-factory/SKILL.md
$CODEX_HOME/skills/product-engineering-ops/SKILL.md
This skill supports three patterns:
product: a product-memory layer built around areas, problems, initiatives, decisions, experiments, metrics, and distilled research.
operations: a Jay Khalife style intelligence system with Knowledge/, Operations/, hub nodes, entity intelligence summaries, and timestamped playbooks.
hybrid: both layers together, with research and frameworks feeding operational or product decisions.
Read references/basb-methodology-map.md when building a Tiago Forte-style product second brain. Read references/runbook.md when you need the operational topology from the PDF. Read references/schemas.md when you need the note types, required properties, and template shapes.
Workflow
1. Inspect the vault and source material first
- Identify whether the vault already exists, is partially structured, or is just a raw folder.
- Inspect the source corpus before editing anything: markdown, PDFs, DOCX, exported help docs, URLs, repos, meeting notes, tickets, or email exports.
- Decide the mode:
product if the user mostly needs synthesis, reasoning, and product judgment.
operations if the user manages recurring real-world entities such as customers, campuses, projects, cases, or portfolios.
hybrid if both are true.
2. Scaffold the vault before heavy synthesis
Use the scaffold script to create the repeatable structure and Obsidian config:
python3 scripts/scaffold_vault.py \
--vault "/absolute/path/to/vault" \
--project "Project Name" \
--mode hybrid \
--entity-singular customer \
--entity-plural customers
The scaffold creates:
- Official/core Obsidian settings for Properties, Daily Notes, Templates, Backlinks, Search, Canvas, Workspaces, and Bases.
- Product-memory zones such as
00 Home, 10 Areas, 20 Problems, 30 Initiatives, 40 Research, 50 Decisions, 60 Experiments, 70 Journal, 80 Assets, and 90 Templates.
- Product BASB zones and hubs for CODE, PARA, intermediate packets, output conversion, reviews, and archives.
- Operations zones such as
Knowledge/, Operations/, and Operations/_nodes/.
CLAUDE.md and TRAVERSAL-INDEX.md seeds so agents and humans have a navigation contract.
- Valid
.base files, templates, and a starter canvas.
When the user already has a vault, preserve useful material and move raw or imported content into the research or knowledge layer instead of deleting it.
For existing vaults that only need the Product BASB surfaces and metadata, use the migration script in dry-run mode first:
python3 scripts/migrate_to_product_basb.py \
--vault "/absolute/path/to/vault" \
--product-slug "product-slug"
Add --write only after reviewing the report. Use --overwrite-generated only for scaffolded or generated hubs/templates.
3. Ingest sources without losing provenance
- Preserve raw material in
40 Research or Knowledge/Books|Frameworks|Concepts|....
- If a source is authenticated or externally linked, capture enough context to make the note usable offline.
- Prefer one note per durable concept, entity, problem, decision, or experiment.
- Tag durable product notes with CODE/PARA metadata:
basb_stage, para_category, distillation_level, actionability, and optional output_target.
- When importing from an existing vault, patch stale queries and legacy home notes after any folder moves.
4. Build the graph deliberately
These rules are non-negotiable:
- Every Markdown file should have YAML frontmatter.
- Every note should have at least one wikilink.
- Hub nodes aggregate; leaf nodes contain.
- Use links for durable nouns and tags for workflow states such as
#inbox, #to-review, #decision-needed, and #risk.
- Never overwrite playbooks or action plans; create timestamped versions.
For product and hybrid mode:
- Create area, problem, initiative, metric, decision, experiment, and insight notes.
- Keep raw research in
40 Research.
- Use
Product OS, CODE Dashboard, PARA Map, Output Pipeline, Areas, Decision Log, and Active Bets as the main dashboards.
- Treat
30 Initiatives as Projects, 10 Areas as Areas, 40 Research as Resources, and 90 Archive as Archives.
For operations and hybrid mode:
- Create one
_intelligence_summary.md per real-world entity.
- Create hub nodes in
Operations/_nodes/ for the dimensions the user actually filters by.
- Use
Knowledge/MOCs/ to bridge knowledge-layer frameworks to operational situations.
- Keep each entity folder as a living intelligence space for emails, transcripts, tickets, and timestamped playbooks.
5. Add code intelligence when repos matter
If the source corpus includes repositories, create a code-intelligence layer adjacent to research:
- Repo notes with purpose, stack, key files, and representative code.
- Product-to-code or operations-to-code maps.
- Architecture deep dives or playbooks for recurring workflows.
- Links from product problems or operational issues back to the relevant repos and files.
This layer belongs in research, not as a replacement for product or operational reasoning.
6. Maintain the vault as a synthesis layer
- Use Daily Notes as the capture inbox.
- Distill raw capture into durable notes at least weekly using progressive summarization.
- Move reusable research, code, support, and planning fragments into intermediate packets that can feed future initiatives.
- Convert high-value intelligence into shippable outputs such as PRDs, specs, tickets, pull requests, runbooks, decisions, launch notes, and post-launch learnings.
- Keep generated output candidates as vault drafts unless an approved delivery-system adapter is explicitly configured.
- Use weekly review and archive records to close the CODE loop without rewriting user-authored initiative notes.
- Keep delivery in the system of record such as GitHub, Jira, or Linear; Obsidian should store the reasoning, evidence, and intelligence around the work.
- Revisit
CLAUDE.md, TRAVERSAL-INDEX.md, dashboards, and hub nodes as the vault evolves.
7. Audit after every substantial change
Run the audit script before finishing:
python3 scripts/audit_vault.py --vault "/absolute/path/to/vault"
Use --write to save a report into the vault:
python3 scripts/audit_vault.py \
--vault "/absolute/path/to/vault" \
--write "/absolute/path/to/vault/80 Assets/vault-audit.md"
The audit checks:
- Missing frontmatter
- Missing wikilinks
- Missing required fields for known note types
- Duplicate stems that can make wikilinks ambiguous
- Orphan candidates with no inbound or outbound graph connections
8. Propagate source-of-truth changes into specialization vaults
When this skill changes, any specialization built on top of it should be treated as stale until its vault is refreshed.
- Treat the skill source of truth as the combination of:
SKILL.md
references/
scripts/
agents/
- Specializations should watch both this general skill directory and their own specialization directory.
- Use a hash-based change detector so local edits and pulled PR merges are both caught once the source-of-truth files land locally.
- Only update the saved watcher state after the specialization refresh pipeline succeeds.
- Prefer an automation for this instead of relying on memory. The automation should run the specialization refresh script, not try to rebuild the vault ad hoc.
9. Prefer manifests over custom per-product workflows when scaling
If the user is trying to onboard many products, this skill should not become the only place where product-specific logic lives.
- Put product identity, sources, repos, automation ids, and readiness categories into a manifest.
- Keep this skill generic.
- Push scaling concerns into the product factory layer instead of copying and pasting one-off product workflows.
What this skill should usually produce
- A structured vault with clear entry points
- A research or knowledge layer that preserves provenance
- A product-memory or operations-intelligence layer that is actually navigable
- Templates and Bases views so the vault stays consistent
- A CLAUDE-style agent schema and traversal index
- An audit report or a final cleanup pass that removes structural blind spots
Example requests
- "Turn these PDFs, docs, and repos into an Obsidian second brain for our product."
- "Create a customer intelligence system in Obsidian for our campus portfolio."
- "Reorganize this vault into research, decisions, initiatives, and experiments without losing the imported notes."
- "Use this support export plus GitHub repos to build a linked product-memory vault."