# Obsidian Link Hub Builder

> Build additive companion hub notes for one bounded Obsidian topic cluster, improving navigation and link structure through derived notes under _codex. Use for companion hubs, not for broad atlases, direct note editing, or project-wide normalization.

- Skill: `kirillklem/obsidian-link-hub-builder` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add kirillklem/obsidian-link-hub-builder`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kirillklem/obsidian-link-hub-builder/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: KirillKlem (https://skillmd.com/u/kirillklem)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/kirillklem/obsidian-link-hub-builder

---


# Obsidian Link Hub Builder

## When to Use
- A cluster has many concept notes but weak atlas coverage
- A dashboard note exists but is too thin
- The user wants better navigation without mass-linking or rewriting source notes
- `_codex/linking_opportunities.md` has already identified a high-confidence target

## When Not to Use
- Do not use for whole-vault mapping; prefer `obsidian-topic-atlas`
- Do not use for extracting formulas or definitions; prefer `obsidian-definition-extractor`
- Do not use for broad project normalization; prefer `obsidian-project-normalizer`
- Do not use when the cluster boundary is still too ambiguous to name

## Inputs
- Bounded topic cluster, e.g. VLM/grounding, RL roadmap, football analytics, interview prep
- Existing atlas or linking audit if available
- Safe write area, usually `learn/_codex/`
- Local vault rules from `learn/AGENTS.md`

## Output Files
- Primary: `_codex/<topic_slug>_hub.md`
- Optional variant: `_codex/<topic_slug>_reading_map.md`

## Constraints
- Do not edit original notes to add backlinks
- Keep the hub bounded to one coherent cluster
- Prefer high-confidence grouping over forced completeness
- Distinguish canonical notes from derived navigation notes
- Do not pull in personal-sensitive notes unless the cluster truly requires them

## Procedure
1. Read `learn/AGENTS.md` and any existing `_codex/topic_atlas.md` or `_codex/linking_opportunities.md`.
2. Pick one high-confidence cluster only.
3. Identify:
   - cluster entry note
   - anchor concept notes
   - leaf notes
   - thin or missing intermediate hubs
4. Read a selective subset of representative notes.
5. Write a derived hub containing:
   - cluster summary
   - core notes
   - sub-branches
   - suggested reading order
   - unresolved or missing nodes
6. Link outward to canonical notes rather than rephrasing all content.
7. Save the hub under `_codex/`.
8. Verify that the new hub reduces navigation cost and does not impersonate the canonical source.

## Definition of Done
- The hub is bounded to one coherent cluster and named clearly
- It identifies core notes, sub-branches, and missing nodes
- It improves traversal cost without pretending to be the source of truth
- It makes ambiguity visible instead of hiding it behind forced structure
- A future agent could use it as a starting point for deeper work in that cluster

## Vault-Specific Guidance
- Strong default targets in `learn` include:
  - VLM / grounding / open-vocabulary vision
  - attention systems
  - AiStats / football analytics
  - RL roadmap
  - interview preparation
- Use the existing `_codex/linking_opportunities.md` file as the first candidate list when it exists.
- Use `references/hub_note_template.md` when you need a repeatable hub format.