Marketplace for trusted AI agent skills

Skill Detail

Turn Agent Corrections Into Durable Skills With claude-smart

Use claude-smart to capture corrections and working patterns as preferences, project skills, and shared skills that Claude Code, Codex, and OpenCode reuse later.

Templates & WorkflowsMulti-Framework
Templates & Workflows Multi-Framework Security Reviewed
⭐ 748 GitHub stars ⬇ 1.9k/wk npm
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill turn-agent-corrections-into-durable-skills-with-claude-smart Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
View source Documentation
At a glance
Tools required
Node.js 20+, npx, target host CLI on PATH such as claude, codex, or opencode; learned local data under ~/.reflexio/ and ~/.claude-smart/
Install & setup
For Claude Code, run npx claude-smart install. For Codex, run npx claude-smart install --host codex. For OpenCode, run npx claude-smart install --host opencode; use --global for all OpenCode projects. Restart the host after installation.
Author
Yi Lu / ReflexioAI
Publisher
Organization
Last updated
Jul 20, 2026
Quick brief

claude-smart is a self-improvement plugin for Claude Code, Codex, and OpenCode. The operator installs it into the host CLI so corrections, successful project patterns, and reusable guidance become durable preferences, project-specific skills, and shared skills for later sessions.

How it works

What this skill actually does

Use this instead of relying on normal chat memory when the agent keeps repeating corrected behavior or when repo-local working practices should survive across restarts and related tools. The workflow is local-first and focused on converting real interaction history into explicit reusable rules rather than replaying whole transcripts.

The scope boundary is the installed claude-smart/reflexio plugin workflow. It is not a generic memory server, a notes app, or a broad agent framework. Its approved use is supervised learning from corrections into auditable preference and skill artifacts for supported coding-agent hosts.

Inputs and prerequisites: Node.js 20+, npx, target host CLI on PATH such as claude, codex, or opencode; learned local data under ~/.reflexio/ and ~/.claude-smart/.

Setup notes: For Claude Code, run npx claude-smart install . For Codex, run npx claude-smart install –host codex . For OpenCode, run npx claude-smart install –host opencode ; use –global for all OpenCode projects. Restart the host after installation.

Source and verification boundary: use https://www.reflexio.ai/docs/claude-smart as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.

Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.