index-learnings

Indexes learning markdown files into SQLite and generates a topic manifest.

1|1|Updated Jan 12, 2026
One-click install
npx skills add https://github.com/Connsulting/claude-code-plugins --skill index-learnings-connsulting
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: index-learnings
Source: https://github.com/Connsulting/claude-code-plugins/tree/main/plugins/compound-learning/skills/index-learnings
Command: npx skills add https://github.com/Connsulting/claude-code-plugins --skill index-learnings-connsulting

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sqlite-vec.

What problem does it solve? Learning notes scattered across global and repository-specific directories are hard to search and summarize. This Skill indexes all learning markdown files into a SQLite database and produces a manifest that summarizes topics, keywords, and gotchas for quick discovery. ## Core Features & Use Cases - Bulk Indexing: Discovers .md files in ~/.projects/learnings/ (global) and [repo]/.projects/learnings/ (repo), extracting **Topic:**, **Tags:**, and **Type:** fields into SQLite. - Manifest Generation: Writes ~/.projects/learnings/MANIFEST.md with per-scope tables of topics, counts, keywords, and gotcha flags. - Orphan Pruning & Stable Dates: Removes database entries whose files no longer exist and preserves original creation dates from metadata, filename date suffixes, or file mtime. - Use Case: After accumulating dozens of learning notes across projects, run the indexer to rebuild the searchable database and regenerate a manifest you can embed in CLAUDE.md for context. ## Quick Start Ask the AI to run /index-learnings to index all learning files and regenerate the topic manifest.

Frequently Asked Questions about index-learnings

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I index markdown learning notes into SQLite?▼

Run the /index-learnings command, which discovers all .md files in the global and repo learnings directories, extracts Topic and Tags fields, and upserts them into SQLite. The database auto-initializes on first run and is safe to re-run.

What format do learning files need for indexing?▼

Learning files are markdown documents with explicit **Type:**, **Topic:**, and **Tags:** fields, followed by Problem and Solution sections. Files without a Topic field fall back to tag-based topic inference or the 'other' category.

Does the indexer support both global and repo-specific learnings?▼

Yes, it scans ~/.projects/learnings/ for global learnings and searches for .projects/learnings directories under the configured repo search path. The generated manifest separates global and per-repo sections.

Why does re-indexing reset my learning creation dates?▼

It does not in this implementation. The indexer derives created_at from an explicit Created/Learned/Date field, then the filename date suffix (name-YYYY-MM-DD.md), then file mtime, so re-indexing preserves the original age signal.

What happens to database entries when learning files are deleted?▼

After indexing, the script checks every stored document's file_path and deletes entries whose files no longer exist on disk. It reports how many orphaned entries were pruned.