rohitg00/pro-workflow

Claude Code learns from your corrections: self-correcting memory that compounds over 50+ sessions. Context engineering, parallel worktrees, agent teams, and 17 battle-tested skills.

What it solves

Pro Workflow addresses the "memory loss" and repetitive corrections common in AI coding assistants like Claude Code and Cursor. It prevents users from having to explain the same conventions or correct the same mistakes across multiple sessions by providing a durable, persistent memory layer.

How it works

The project implements a single SQLite store that acts as a shared memory and knowledge base for AI agents. It uses FTS5 (Full-Text Search) for fast retrieval of rules and wiki pages. The system operates through three primary mechanisms:

  1. Self-Correction Memory: Corrections made by the user are captured as rules and automatically loaded at the start of every new session.
  2. Knowledge Plane: Users can create persistent research wikis that the agent can query, expand via an auto-research loop (BFS), and auto-inject into prompts when relevant topics are mentioned.
  3. Agentic Orchestration: A library of 41 skills, 8 specialized agents (e.g., Planner, Reviewer, Debugger), and 23 slash commands that structure the development process into a Research $\rightarrow$ Plan $\rightarrow$ Implement $\rightarrow$ Review pipeline with validation gates.

Who it’s for

Developers using AI coding agents—specifically Claude Code, Cursor, and other CLI-based agents—who want to maintain a long-term knowledge base and enforce consistent coding standards across sessions.

Highlights

  • Self-Correcting Memory: Automatically turns user corrections into searchable rules to eliminate repetitive mistakes.
  • Auto-Research Loop: A budget-capped, autonomous loop that grows research wikis using web, arXiv, and GitHub fetchers.
  • Cross-Agent Compatibility: Works with over 32 agents (including Cursor and Gemini CLI) via a unified skill installer.
  • Hybrid Retrieval: Combines BM25 and vector search (RRF) for high-accuracy knowledge retrieval.
  • Quality Gates: Includes LLM-powered hooks for secret scanning, commit validation, and state preservation during context compaction.