skills.rest
surfingalien
surfingalienCOMMUNITY
Explore all skills by this author →
GitHub

pro-workflow

Implements AI coding workflow patterns including self-correction loops, hooks, agents, and orchestration for Claude Code and Cursor.

1|Updated May 6, 2026
Search Tags:#claude code#hooks#subagents#orchestration#ai workflow#claude.md#cursor
One-click install
npx skills add https://github.com/surfingalien/FinSurfing --skill pro-workflow-surfingalien
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pro-workflow
Source: https://github.com/surfingalien/FinSurfing/tree/main/.claude/skills/pro-workflow
Command: npx skills add https://github.com/surfingalien/FinSurfing --skill pro-workflow-surfingalien

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI coding assistants repeat mistakes, lose context, and require constant correction. This Skill provides a structured workflow system that captures corrections as persistent rules, enforces quality gates through hooks, and orchestrates multi-phase development so AI-generated code requires less manual rework. ## Core Features & Use Cases - Self-Correction Loop: Converts user corrections into persistent LEARNED rules in CLAUDE.md so the same mistake never repeats across sessions. - Hooks & Quality Gates: 18 hook events (PreToolUse, PostToolUse, SessionStart, Stop, SessionEnd) that remind about linting, testing, and uncommitted changes at checkpoints. - Orchestration Patterns: Command > Agent > Skill architecture with a four-phase /develop workflow (research, plan, implement, review) for features touching more than five files. - Cross-Agent Support: Works with Claude Code, Cursor, Codex, Gemini CLI, and 32+ agents via SkillKit translation. - Use Case: A developer notices Claude editing the wrong file repeatedly; the self-correction protocol proposes a rule, stores it in LEARNED, and future sessions confirm full paths before editing. ## Quick Start Add the self-correction protocol and pre-flight discipline sections to your CLAUDE.md file, then ask the AI to propose a LEARNED rule the next time you correct it.

Frequently Asked Questions about pro-workflow

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

FAQPage Schema
How do I make Claude Code remember my corrections across sessions?▼

Use the self-correction protocol: when you correct the AI, it proposes a one-line rule prefixed with [LEARN], and after your approval the rule is appended to the LEARNED section of CLAUDE.md. Future sessions load these rules automatically.

How to set up hooks in Claude Code for quality gates?▼

Copy the provided hooks.json into ~/.claude/settings.local.json or merge it with existing settings. Hooks fire on PreToolUse, PostToolUse, and session events to remind you about linting, testing, and uncommitted changes at checkpoints.

Does this workflow system work with Cursor or only Claude Code?▼

It works with Cursor, Codex, Gemini CLI, and 32+ agents. Claude Code-specific features like hooks use settings files, while Cursor users get equivalent enforcement through .mdc rules with alwaysApply enabled, and SkillKit translates skills between agents.

What is the difference between commands, agents, and skills in Claude Code?▼

Commands are user-facing entry points, agents execute with constrained tools and preloaded skills, and skills provide domain knowledge injected at startup. The orchestration pattern layers them so a command like /develop coordinates agents through research, plan, implement, and review phases.

When should I use git worktrees for parallel AI coding sessions?▼

Use worktrees when blocked on long operations like test runs or builds, or when exploring multiple approaches simultaneously. Run claude -w for an isolated worktree session, or git worktree add manually in any editor, so parallel work never conflicts.

Why does my Claude Code session run out of context so quickly?▼

Context fills from verbose tool output, unused MCP servers, and long explorations. Keep CLAUDE.md under 60 lines, enable fewer than 10 MCPs, delegate heavy exploration to subagents, and set CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=50 for earlier compaction.

Related Skills

Explore More Software Engineering Skills→
d0whc3rd0whc3r
2

speckit-git-initialize

Initialize a Git repository and create an initial commit.

Community
Intermediate
trust-chain-organizationtrust-chain-organization

project-conventions

Enforce Polibase project conventions during commits, PRs, and CI/CD failures.

Official
Intermediate
Till-Crazy-Tears-Us-ApartTill-Crazy-Tears-Us-Apart
27

tool-guide

Directs users to the appropriate Claude Code tool for searches, docs, or browser automation.

Community
Basic
ab180ab180

day1-onboarding

Onboard learners to Claude Code installation and seven core AI-native workflow capabilities.

Official
Intermediate