feat: Continual Learning hooks, skill, and docs site update

[thegovind]

What

Adds continual learning infrastructure for AI coding agents — a single hook that enables agents to learn, reflect, and persist knowledge across sessions.

Inspired by the invariants in beads: one-step setup, auto-initialization, structured data with compaction.

Install

cp -r hooks/continual-learning .github/hooks/

That's it. Auto-initializes on first session.

Design

Two-Tier Memory

Scope	Location	Purpose
Global	~/.copilot/learnings.db	Tool patterns, cross-project insights — follows you everywhere
Local	.copilot-memory/learnings.db	Repo conventions, project-specific mistakes — stays with the project

Single Script

One learn.sh handles all lifecycle events — no multi-script complexity:

• sessionStart → loads learnings from both DBs
• postToolUse → tracks tool outcomes (3ms overhead)
• sessionEnd → analyzes patterns, persists insights, compacts old data

Compaction

Old low-value learnings decay automatically (60-day TTL, low hit count). High-value learnings persist and rank higher. Tool logs pruned after 7 days.

Changes

🧠 Hook (hooks/continual-learning/)

• learn.sh — single self-contained script, auto-inits DBs and directories
• hooks.json — lifecycle event configuration
• README.md — concise docs

📚 Skill (.github/skills/continual-learning/)

• Teaches the continual learning pattern
• Covers hook-based, agent-native (store_memory/SQL), and memory file approaches

🌐 Docs Site

• New Hooks section with featured card (purple gradient, 🧠)
• HookCard.astro component
• Subtle link to awesome-copilot for more hooks
• Synced skills data (133 → 134 skills)