# Obsidian Definition Extractor

> Extract definitions, formulas, and compact glossaries from concept-heavy Obsidian notes into derived companion notes under _codex. Use for term extraction and formula sheets, not for full note rewriting or broad topic atlases.

- Skill: `kirillklem/obsidian-definition-extractor` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add kirillklem/obsidian-definition-extractor`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kirillklem/obsidian-definition-extractor/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-definition-extractor

---


# Obsidian Definition Extractor

## When to Use
- A note is dense and future agents need a cleaner definitions layer
- The user wants formulas, concepts, or short canonical meanings extracted from several related notes
- A cluster has many leaf notes but no compact glossary or reference sheet

## When Not to Use
- Do not use for broad thematic mapping; prefer `obsidian-topic-atlas`
- Do not use for project-structure cleanup; prefer `obsidian-project-normalizer`
- Do not use when the source material is mostly narrative or personal reflection rather than concept-heavy content
- Do not use when the task requires editing the source notes directly

## Inputs
- Source note or bounded note set, e.g. `Knowledge/Obj-detection.md` plus related detection notes
- Target term scope, e.g. "DETR family", "normalization layers", "RL core terms"
- Safe write area, usually `learn/_codex/`
- Existing vault guidance from `learn/AGENTS.md`

## Output Files
- Single-note extraction: `_codex/<topic_slug>_definitions.md`
- Cross-note glossary: `_codex/<topic_slug>_glossary.md`
- Formula-focused variant: `_codex/<topic_slug>_formula_sheet.md`

## Constraints
- Do not alter original notes to insert extracted definitions
- Do not silently merge unrelated topics into one glossary
- Preserve uncertainty when the source note is informal or incomplete
- Quote minimally; prefer paraphrase and explicit links back to the source notes
- Keep the run bounded to one topic family at a time

## Procedure
1. Read `learn/AGENTS.md` and confirm `_codex/` is the write-safe area.
2. Select one topic cluster with clear boundaries.
3. Read the smallest sufficient set of source notes.
4. Extract only high-confidence items:
   - explicit definitions
   - formulas
   - short method characterizations
   - contrasts between closely related methods
5. Normalize terminology carefully:
   - one heading per term
   - short definition
   - optional "source notes" list
   - optional "related terms" section
6. Separate confirmed definitions from inferred summaries.
7. Write a derived glossary under `_codex/`.
8. Include backlinks to the original source notes rather than copying whole passages.
9. Verify that the file improves retrieval and study, not just duplication.

## Definition of Done
- Every extracted term is tied to an explicit source note or clearly marked as inferred synthesis
- The output is bounded to one concept family instead of becoming a mini-atlas
- Formulas and definitions are separated cleanly when both appear
- The new file improves lookup speed for future agents and humans
- The result contains minimal quotation and no silent mutation of canonical notes

## Vault-Specific Guidance
- Good candidates in `learn` include:
  - detection metrics and evaluation
  - transformer attention variants
  - RL methods
  - normalization / activation families
  - football analytics metrics
- Avoid extracting from highly personal notes unless the user explicitly asks.
- Use `references/definition_output_template.md` if you want a stable layout for term pages.