Skillget

flow-next

agent
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 560 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose: This tool provides a suite of workflow plugins designed for AI coding assistants like Claude Code and OpenAI Codex. It focuses on plan-first development, autonomous overnight coding, and utilizing multiple AI models for code review gates.

Security Assessment: Overall risk is Low. The automated code scan checked 12 files and found no dangerous patterns, hardcoded secrets, or requests for dangerous permissions. The tool relies heavily on executing shell commands and making external network requests, which is standard for its intended purpose (interacting with local git repositories and external AI models). However, because it facilitates autonomous actions like "overnight coding," developers should be aware of what commands the agent is executing in their environment without real-time supervision.

Quality Assessment: High. The project is actively maintained (last pushed today) and features a highly permissive MIT license, making it safe for commercial and personal use. It has strong community trust, evidenced by 560 GitHub stars, and provides clear documentation, Discord support, and dedicated install scripts for cross-platform compatibility.

Verdict: Safe to use.
SUMMARY

Claude Code plugins for reliable AI coding. Flow-Next: plan-first workflows, Ralph autonomous mode (overnight coding with fresh context), multi-model review gates via RepoPrompt/Codex, re-anchoring to prevent drift, receipt-based gating.

README.md

gmickel claude marketplace

License: MIT
Claude Code

Flow-next
Flow-next Docs

Author
Twitter
Sponsor
Discord

Plugins that make AI agents actually work.

💬 Join the Discord — discussions, updates, feature requests, bug reports

🔄 Update issues? Run: claude plugin update flow-next@gmickel-claude-marketplace

🆕 /flow-next:epic-review: Epic-completion review gate. Catches requirement gaps that per-task reviews miss—before epic closes.

🤖 Ralph mode: Ship features while you sleep. Fresh context per iteration, multi-model review gates, auto-blocks stuck tasks.

📡 Cross-platform reviews: RepoPrompt (macOS) or Codex CLI (any OS) · Codex install

🧪 OpenCode user? Try flow-next-opencode (experimental port)

[!TIP]
🤖 Works on Factory Droid too! Claude Code plugins are fully compatible:

droid plugin marketplace add https://github.com/gmickel/gmickel-claude-marketplace

Then /plugins → Marketplace → install flow-next. Commands don't autocomplete yet but work when typed (e.g. /flow-next:plan). Skills load automatically.

[!TIP]
🖥️ OpenAI Codex user? Full multi-agent support (20 roles, parallel scouts, 3-tier model mapping):

git clone https://github.com/gmickel/gmickel-claude-marketplace.git
cd gmickel-claude-marketplace && ./scripts/install-codex.sh flow-next

Requires Codex 0.102.0+. Commands use /prompts: prefix (e.g. /prompts:plan). See Codex install guide.

The Problem

Process failures, not model failures:

  • Starting to code before understanding the codebase
  • Reinventing patterns already there
  • Forgetting the plan mid-implementation
  • Skipping edge cases obvious in hindsight

This marketplace ships plugins that fix these problems.

Flow-Next

Plan first, work second. Zero external dependencies.

# Install
/plugin marketplace add https://github.com/gmickel/gmickel-claude-marketplace
/plugin install flow-next

# Setup (configures review backend + CLI tools)
/flow-next:setup

# Use
/flow-next:plan Add a contact form with validation
/flow-next:work fn-1

Why It Works

Problem Solution
Context drift Re-anchoring before EVERY task — re-reads specs + git state from .flow/
200K token limits Fresh context per task — worker subagent starts clean each task
Single-model blind spots Cross-model reviews — RepoPrompt or Codex as second opinion
Forgotten requirements Dependency graphs — tasks declare blockers, nothing runs out of order
"It worked on my machine" Evidence recording — commits, test output, PRs tracked per task
Infinite retry loops Auto-block stuck tasks — fails after N attempts, moves on
Team conflicts Multi-user safe — scan-based IDs, soft claims, no coordination server

Commands

Command What It Does
/flow-next:plan Research codebase, create epic + tasks
/flow-next:work Execute tasks with re-anchoring
/flow-next:interview Deep spec refinement (40+ questions)
/flow-next:plan-review Cross-model plan review
/flow-next:impl-review Cross-model implementation review
/flow-next:prime Assess codebase agent-readiness, propose fixes
/flow-next:ralph-init Scaffold autonomous loop

📖 Full documentation — CLI reference, workflow details, troubleshooting

🤔 Confused when to use Interview vs Plan vs Work? See When to Use What

Ralph (Autonomous Mode)

Run overnight, walk away. Fresh context per iteration + multi-model review gates.

/flow-next:ralph-init           # One-time setup
scripts/ralph/ralph.sh          # Run from terminal

How Ralph differs:

Aspect Typical Agents Ralph
Context Accumulates (drift) Fresh each iteration
Review Self-review only Cross-model gates
Stuck tasks Infinite retry Auto-block after N failures
Validation Tests only Tests + receipts + reviews

📖 Ralph deep dive — guard hooks, receipt gating, sentinel controls

🖥️ Ralph TUI — Terminal UI for monitoring (bun add -g @gmickel/flow-next-tui)

Install

# Add marketplace
/plugin marketplace add https://github.com/gmickel/gmickel-claude-marketplace

# Install flow-next
/plugin install flow-next

# Setup (configures review backend + CLI tools)
/flow-next:setup

Other Plugins

Plugin Status
flow-next Active development. Recommended.
flow Legacy. Documentation

Ecosystem

Project Platform
flow-next-opencode OpenCode
FlowFactory Factory.ai Droid

Also Check Out

GNO — Local hybrid search for your notes, docs, and code. Give Claude Code long-term memory over your files via MCP.

bun install -g @gmickel/gno && gno mcp install --target claude-code

Contributing

  1. Create plugins/<name>/ with .claude-plugin/plugin.json
  2. Add commands/agents/skills under that plugin root
  3. Update .claude-plugin/marketplace.json
  4. Validate: jq . .claude-plugin/marketplace.json

Yorumlar (0)

Sonuc bulunamadi