remember-learnings

Extracts session learnings and updates AGENTS.md and rules files with actionable guidance.

21.4k|2.6k|Updated Apr 11, 2025
One-click install
npx skills add https://github.com/dyad-sh/dyad --skill remember-learnings
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: remember-learnings
Source: https://github.com/dyad-sh/dyad/tree/main/.claude/skills/remember-learnings
Command: npx skills add https://github.com/dyad-sh/dyad --skill remember-learnings

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Hard-won knowledge from a coding session—errors fixed, gotchas discovered, workflow friction resolved—is usually lost when the session ends, forcing future agent sessions to repeat the same mistakes.

Core Features & Use Cases

  • Session Analysis: Reviews the conversation history for errors, snags, workflow friction, and undocumented architecture insights.
  • Targeted Documentation Updates: Places each learning into the most relevant rules/*.md topic file, creates new rule files when warranted, and updates the rules index in AGENTS.md.
  • Maintenance and Deduplication: Consolidates overlapping entries, removes obsolete learnings, and limits additions to five high-impact items per session.
  • Use Case: After a long debugging session involving E2E test failures and IPC quirks, run this Skill to record the fixes in rules/e2e-testing.md and rules/electron-ipc.md so the next session avoids the same pitfalls.

Quick Start

Review this session for errors and hard-won knowledge, then update the rules files or AGENTS.md with concise actionable learnings and stage the changes.

Frequently Asked Questions about remember-learnings

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

FAQPage Schema
How do I preserve learnings from an AI coding session?

Run the remember-learnings Skill at the end of a session. It reviews the conversation for errors, snags, and undocumented conventions, then writes concise actionable bullets into the relevant rules/*.md files or AGENTS.md and stages the changes with git.

How to document agent conventions in AGENTS.md and rules files?

Keep AGENTS.md as the top-level guide with a rules index table, and place topic-specific learnings in rules/*.md files such as rules/e2e-testing.md. New topics get their own kebab-case file plus an index entry in AGENTS.md.

Does this Skill edit CLAUDE.md directly?

No. CLAUDE.md is a symlink to AGENTS.md, so all edits must be made to AGENTS.md. Editing CLAUDE.md directly is explicitly prohibited by the Skill instructions.

What happens if a session has no new learnings?

The Skill skips the edit and reports that no updates were needed. It avoids duplicating anything already documented in AGENTS.md or the existing rules files.

How many learnings can be added per session?

At most five learnings per session. Each learning must be one to three lines, specific and actionable, and overlapping or superseded entries are consolidated rather than appended.