lore-knowledge

name	lore-knowledge
description	Query project-specific knowledge for conventions, patterns, architecture decisions, gotchas, and domain context. Use when starting a new task, touching unfamiliar code, making architectural decisions, reviewing code patterns, or needing context about how things are done in this project.
allowed-tools	Bash, Read

Lore Knowledge System

This skill provides access to the project's Lore knowledge database - a curated collection of conventions, patterns, architecture decisions, gotchas, and domain knowledge specific to this codebase.

Before Querying

First, check if Lore is available in this project:

lore doctor

If Lore is not initialized (no .lore/ directory), inform the user that this project doesn't have Lore set up yet. They can initialize it with lore init.

Querying Knowledge

By Task (Most Common)

When starting a new task or feature:

lore query --task="<describe what you're working on>"

Examples:

• lore query --task="implement user authentication"
• lore query --task="add API endpoint for payments"
• lore query --task="refactor database queries"

By Files

When working with specific files or directories:

lore query --files="<glob pattern>"

Examples:

• lore query --files="src/api/**"
• lore query --files="packages/auth/*"
• lore query --files="src/components/Button.tsx"

By Tags

When looking for specific categories of knowledge:

lore query --tags="<tag1>,<tag2>"`

Examples:

• lore query --tags="security"
• lore query --tags="testing,api"
• lore query --tags="performance"

Combined Queries

You can combine parameters for more targeted results:

lore query --task="add validation" --files="src/api/**" --tags="security"

Understanding Results

Query results include:

• Title: Brief description of the knowledge
• Type: convention, pattern, architecture, decision, gotcha, workflow, reference, or example
• Relevance Score: How well it matches your query (higher is better)
• Freshness: How recently the knowledge was updated (hot/warm/cold/frozen)
• Content: The actual knowledge entry

Prioritize entries with:

1. Higher relevance scores
2. "Hot" or "warm" freshness
3. Types matching your need (e.g., "gotcha" for pitfalls, "pattern" for code patterns)

When Results Are Empty

If no knowledge is found:

1. The project may have limited knowledge captured
2. Try broader search terms
3. Query by file path if working on specific code
4. Suggest the user capture relevant knowledge after discovering it

Capturing New Knowledge

When you discover something worth preserving (a pattern, gotcha, decision, etc.):

lore add --title="<concise title>" --type=<type> --body="<markdown content>"

Types available:

• convention - How things are done here
• pattern - Recurring code patterns
• architecture - System design decisions
• decision - ADR-style decisions with context
• gotcha - Pitfalls and common mistakes
• workflow - Process documentation
• reference - Quick lookup information
• example - Code samples and usage examples

Example:

lore add --title="API Error Response Format" --type=convention --body="All API errors must return JSON with {error: string, code: number, details?: object}"

Quick Reference

Action	Command
Check setup	lore doctor
Query by task	lore query --task="..."
Query by files	lore query --files="..."
Query by tags	lore query --tags="..."
List all entries	lore list
Show entry details	lore show <id>
Capture knowledge	lore add --title="..." --type=... --body="..."
Get full context	lore prime