</>MCP Agents Market
Skill

self-learning-skills

by Kulaxyz951Updated 2026-07-01

A self-improving skill for AI coding agents (Claude Code, Cursor, AGENTS.md): recognize a hard-won golden path in a session and harvest it into a reusable skill/rule for next time.

Claude CodeCursorCodexClineOpenCodeZedAiderGemini CLI

Self-learning-skills is a meta-skill for AI coding agents that automatically captures and persists golden paths discovered during debugging sessions. When an agent solves a non-obvious problem—finding production credentials, discovering the correct deploy command, or debugging through multiple failed attempts—this skill recognizes the breakthrough moment and saves the procedure for automatic reuse in future sessions. It works with Claude Code, Cursor, Codex, Zed, Aider, and any agent that reads AGENTS.md files, storing captured workflows as new skills, rules, or memory notes depending on complexity and verification confidence.

Key Features

Automatically recognizes reusable golden paths during agent sessions without manual prompting
Captures multi-step procedures including what didn't work to skip dead-ends in future sessions
Adapts storage format per tool: skills/ for Claude Code, .cursor/rules/ for Cursor, AGENTS.md for generic agents
Applies promotion rules requiring verified success, named failure patterns, and ruled-out dead-ends before creating skills
Routes lessons by granularity: procedures become skills, facts go to lightweight memory, one-offs are skipped
Built-in secret safety that records only where to find credentials, never the values themselves
Installs via npx skills command for 70+ agents or manually via git clone
Built on the open Agent Skills standard with templates for consistent skill authoring

Use Cases

  • 01Preserving hard-won database connection procedures across debugging sessions
  • 02Remembering project-specific deploy commands and verification workflows discovered through trial and error
  • 03Building a library of operational workflows that recur across sprints
  • 04Capturing diagnostic procedures that worked after multiple failed attempts
  • 05Avoiding repeated exploration of known dead-end approaches in complex debugging
  • 06Accumulating project-specific knowledge that new team members' agents can auto-load

Related Skills

View more

self-learning-skills — FAQ

What is self-learning-skills?+

Self-learning-skills is a meta-skill for AI coding agents that automatically recognizes when you've discovered a reusable debugging procedure or workflow and saves it so the next session starts with that knowledge already loaded. It works by detecting breakthrough moments—commands that only worked after several tries, non-obvious procedures, or when you explicitly say 'remember this'—and persisting them to the appropriate format for your agent.

How do I install self-learning-skills?+

The recommended method is npx skills add kulaxyz/self-learning-skills, which auto-detects and installs for Claude Code, Cursor, Codex, Cline, and 70+ other agents. For Claude Code only, use /plugin marketplace add kulaxyz/self-learning-skills then /plugin install self-learning@self-learning-skills. Manual installation involves copying files to ~/.claude/skills/, .cursor/rules/, or appending to AGENTS.md depending on your agent.

Which AI agents and clients work with self-learning-skills?+

Self-learning-skills supports Claude Code, Cursor, Codex, Cline, OpenCode, Zed, Aider, Gemini CLI, and any agent that reads AGENTS.md or standing-instructions files. The skill adapts its storage format per tool: new skills in skills/ for Claude Code, rules in .cursor/rules/learned/ for Cursor, and appended entries in AGENTS.md for generic agents.

Does self-learning-skills require API keys or credentials?+

No API keys or external credentials are required. The skill operates entirely within your existing agent setup and has built-in secret safety to never write passwords, tokens, or API keys into harvested skills—it only records where to find secrets, not the values themselves.

Is self-learning-skills free to use?+

Yes, self-learning-skills is open source and released under the MIT license, making it free for both personal and commercial use.

What gets saved as a skill versus a memory note?+

Multi-step reusable procedures that passed verification, have named failure patterns, and at least one ruled-out dead-end become full skills. Single facts or one-line corrections go to lightweight memory notes like MEMORY.md. Genuine one-offs are skipped entirely to avoid config bloat.

How do I install self-learning-skills?+

Open the source repository on GitHub and follow its README. self-learning-skills is a skill — MCP Agents Market links you directly to the official repo.

Is self-learning-skills free?+

self-learning-skills is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches