continual-learning

Synced v1.1.0 License: MIT Source: microsoft/skills/.github/skills/continual-learning Commit: cde32f9 Latest included change: 2026-02-24

Guide for implementing continual learning in AI coding agents — hooks, memory scoping, reflection patterns. Use when setting up learning infrastructure for agents.

Summary

Purpose: It helps agents carry useful experience across sessions by capturing, reflecting on, persisting, and applying learnings.
When to use: Use it when an agent should adapt to past tool outcomes or retain reusable conventions across sessions while keeping personal and project knowledge separate.
Outputs: It describes a hook-driven learning loop, global and repository-specific memory, direct or manual recording methods, and retention rules that prune older data.

Upstream changes (2, latest included 2026-02-24)

6737dcb 2026-02-24 refactor: simplify continual learning hook — single script, two-tier memory, auto-init Reworked the skill guide around auto-initialized hooks, global and local SQLite memory, agent-native and manual learning storage, and compaction.

1ea0f8b 2026-02-24 feat: add continual learning hooks, skill, and docs site updates Added a continual-learning guide describing the capture, reflect, persist, and apply loop, hook-based and agent-native implementations, memory-file and diary patterns, and best practices.

npx skills add "lettucebo/Skills#v2.2.0@continual-learning" --full-depth

⚠ Available after v2.2.0 is published

Continual Learning for AI Coding Agents

Your agent forgets everything between sessions. Continual learning fixes that.

The Loop

Experience → Capture → Reflect → Persist → Apply
     ↑                                       │
     └───────────────────────────────────────┘

Quick Start

Install the hook (one step):

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

Auto-initializes on first session. No config needed.

Two-Tier Memory

Global (~/.copilot/learnings.db) — follows you across all projects:

Tool patterns (which tools fail, which work)
Cross-project conventions
General coding preferences

Local (.copilot-memory/learnings.db) — stays with this repo:

Project-specific conventions
Common mistakes for this codebase
Team preferences

How Learnings Get Stored

Automatic (via hooks)

The hook observes tool outcomes and detects failure patterns:

Session 1: bash tool fails 4 times → learning stored: "bash frequently fails"
Session 2: hook surfaces that learning at start → agent adjusts approach

Agent-native (via store_memory / SQL)

The agent can write learnings directly:

INSERT INTO learnings (scope, category, content, source)
VALUES ('local', 'convention', 'This project uses Result<T> not exceptions', 'user_correction');

Categories: pattern, mistake, preference, tool_insight

Manual (memory files)

For human-readable, version-controlled knowledge:

# .copilot-memory/conventions.md
- Use DefaultAzureCredential for all Azure auth
- Parameter is semantic_configuration_name=, not semantic_configuration=

Compaction

Learnings decay over time:

Entries older than 60 days with low hit count are pruned
High-value learnings (frequently referenced) persist indefinitely
Tool logs are pruned after 7 days

This prevents unbounded growth while preserving what matters.

Best Practices

One step to install — if it takes more than cp -r, it won't get adopted
Scope correctly — global for tool patterns, local for project conventions
Be specific — "Use semantic_configuration_name=" beats "use the right parameter"
Let it compound — small improvements per session create exponential gains over weeks