agentflow

分类: 工具与效率 | 上传者: kittors | 下载: 0 | 版本: v1.0（最新）

AgentFlow — Multi-CLI agent workflow system with routing, safety detection, sub-agent orchestration, and knowledge graph memory.

更新日志: Source: GitHub https://github.com/kittors/AgentFlow

目录结构

当前层级: 根目录

📁 .github/
- 📁 workflows/
  - 📄 ci.yml 1.0 KB
  - 📄 release.yml 7.0 KB
📁 agentflow/
- 📁 agents/
  - 📄 architect.toml 607 B
  - 📄 codex_agents.toml 430 B
  - 📄 reviewer.toml 580 B
- 📁 core/
  - 📄 acceptance.md 843 B
  - 📄 attention.md 1.4 KB
  - 📄 common.md 2.3 KB
  - 📄 hooks.md 469 B
  - 📄 hooks_claude.md 310 B
  - 📄 hooks_codex.md 373 B
  - 📄 hooks_other.md 354 B
  - 📄 module_loading.md 1.6 KB
  - 📄 subagent.md 5.2 KB
  - 📄 subagent_claude.md 903 B
  - 📄 subagent_codex.md 1.8 KB
  - 📄 subagent_gemini.md 541 B
  - 📄 subagent_opencode.md 463 B
  - 📄 subagent_other.md 546 B
- 📁 functions/
  - 📄 auto.md 193 B
  - 📄 conventions.md 378 B
  - 📄 dashboard.md 326 B
  - 📄 exec.md 2.6 KB
  - 📄 graph.md 353 B
  - 📄 help.md 1.2 KB
  - 📄 init.md 4.0 KB
  - 📄 memory.md 252 B
  - 📄 plan.md 3.8 KB
  - 📄 review.md 306 B
  - 📄 rlm.md 295 B
  - 📄 scan.md 323 B
  - 📄 status.md 1.1 KB
  - 📄 validatekb.md 297 B
- 📁 hooks/
  - 📄 claude_hooks.json 1.2 KB
  - 📄 codex_hooks.toml 436 B
- 📁 project_rules/
  - 📄 ide_cline_agentflow.md 269 B
  - 📄 ide_cursor_agentflow.mdc 633 B
  - 📄 ide_trae_agentflow.md 343 B
  - 📄 ide_vscode_copilot_agentflow.md 382 B
  - 📄 ide_windsurf_agentflow.md 310 B
- 📁 rlm/
  - 📁 roles/
    
    📄 architect.md 786 B
    
    📄 kb_keeper.md 452 B
    
    📄 pkg_keeper.md 283 B
    
    📄 reviewer.md 826 B
    
    📄 synthesizer.md 636 B
    
    📄 writer.md 260 B
- 📁 rules/
  - 📄 cache.md 416 B
  - 📄 scaling.md 660 B
  - 📄 state.md 773 B
  - 📄 tools.md 843 B
- 📁 services/
  - 📄 attention.md 1.1 KB
  - 📄 knowledge.md 1.9 KB
  - 📄 memory.md 2.8 KB
  - 📄 package.md 1.1 KB
  - 📄 support.md 487 B
- 📁 stages/
  - 📄 design.md 3.3 KB
  - 📄 develop.md 4.3 KB
- 📁 templates/
  - 📄 backend_project.md 1.6 KB
  - 📄 changelog_template.md 585 B
  - 📄 conventions_template.md 1.4 KB
  - 📄 frontend_project.md 1.3 KB
  - 📄 fullstack_project.md 1.7 KB
  - 📄 graph_template.md 1.7 KB
  - 📄 kb_templates.md 943 B
  - 📄 plan_templates.md 1.2 KB
  - 📄 python_project.md 1.5 KB
  - 📄 session_summary_template.md 1.1 KB
- 📄 _AGENTS.md 20.6 KB
- 📄 _SKILL.md 2.6 KB
📁 bin/
- 📄 agentflow.js 3.7 KB
📁 cmd/
- 📁 agentflow/
  - 📄 main.go 179 B
📁 docs/
- 📁 troubleshooting/
  - 📄 windows.md 2.4 KB
- 📄 ai-tools-skill-mcp-guide.md 20.8 KB
📁 internal/
- 📁 app/
  - 📄 app.go 25.2 KB
  - 📄 app_test.go 1.3 KB
  - 📄 automation.go 6.7 KB
  - 📄 automation_test.go 5.4 KB
  - 📄 ecosystem.go 13.7 KB
  - 📄 interactive_test.go 931 B
  - 📄 mcp.go 3.8 KB
  - 📄 project_rules_panels.go 2.9 KB
  - 📄 rules.go 4.9 KB
  - 📄 skill.go 3.1 KB
  - 📄 workspace.go 5.1 KB
- 📁 buildinfo/
  - 📄 buildinfo.go 1.5 KB
  - 📄 buildinfo_test.go 1.1 KB
- 📁 config/
  - 📄 constants.go 387 B
  - 📄 files.go 2.4 KB
  - 📄 files_test.go 837 B
  - 📄 targets.go 1.1 KB
- 📁 i18n/
  - 📄 locale.go 2.4 KB
  - 📄 locale_test.go 1.4 KB
- 📁 install/
  - 📄 bootstrap.go 20.1 KB
  - 📄 bootstrap_test.go 4.0 KB
  - 📄 cli_uninstall.go 6.7 KB
  - 📄 cli_uninstall_test.go 4.2 KB
  - 📄 install.go 21.5 KB
  - 📄 install_test.go 9.7 KB
- 📁 kb/
  - 📄 session.go 4.2 KB
  - 📄 session_test.go 1.0 KB
  - 📄 sync.go 4.1 KB
  - 📄 sync_test.go 2.7 KB
  - 📄 template.go 4.6 KB
  - 📄 template_test.go 563 B
- 📁 mcp/
  - 📄 manager.go 14.8 KB
  - 📄 manager_test.go 11.0 KB
  - 📄 market.go 1.7 KB
  - 📄 registry.go 2.2 KB
- 📁 projectroot/
  - 📄 find.go 499 B
  - 📄 root.go 1.1 KB
- 📁 projectrules/
  - 📄 manager.go 8.8 KB
  - 📄 manager_test.go 1.6 KB
  - 📄 targets.go 1.4 KB
- 📁 scan/
  - 📄 arch.go 7.9 KB
  - 📄 convention.go 8.0 KB
  - 📄 convention_test.go 2.4 KB
  - 📄 dashboard.go 3.0 KB
  - 📄 dashboard_test.go 992 B
  - 📄 graph.go 11.1 KB
  - 📄 graph_test.go 2.9 KB
- 📁 skill/
  - 📄 copy.go 1.2 KB
  - 📄 github.go 2.9 KB
  - 📄 manager.go 3.9 KB
  - 📄 skilldir.go 1.6 KB
  - 📄 skillmd.go 2.2 KB
  - 📄 skillsdotsh.go 1.4 KB
  - 📄 skillsdotsh_test.go 504 B
  - 📄 unzip.go 2.3 KB
  - 📄 validate.go 662 B
  - 📄 validate_test.go 617 B
- 📁 targets/
  - 📄 mcp_targets.go 726 B
  - 📄 profiles.go 727 B
  - 📄 targets.go 3.9 KB
  - 📄 targets_test.go 629 B
- 📁 ui/
  - 📄 interactive_flow.go 81.3 KB
  - 📄 interactive_flow_test.go 23.1 KB
  - 📄 ui.go 24.0 KB
  - 📄 ui_test.go 7.5 KB
- 📁 update/
  - 📄 update.go 9.4 KB
  - 📄 update_test.go 13.0 KB
📄 .env.example 416 B
📄 .gitignore 404 B
📄 .pre-commit-config.yaml 288 B
📄 AGENTS.md 20.2 KB
📄 CHANGELOG.md 1.7 KB
📄 CLAUDE.md 1.5 KB
📄 CONTRIBUTING.md 3.1 KB
📄 embed.go 359 B
📄 go.mod 1.2 KB
📄 go.sum 4.2 KB
📄 install.ps1 4.5 KB
📄 install.sh 9.7 KB
📄 LICENSE 1.1 KB
📄 package.json 865 B
📄 README.md 10.1 KB
📄 README_CN.md 9.9 KB
📄 SKILL.md 2.6 KB

SKILL.md

---
name: agentflow
description: "AgentFlow — Multi-CLI agent workflow system with routing, safety detection, sub-agent orchestration, and knowledge graph memory."
version: "1.0.0"
tags:
  - workflow
  - routing
  - safety
  - memory
  - knowledge-graph
  - convention-check
  - architecture-scan
commands:
  - "~init"
  - "~auto"
  - "~plan"
  - "~exec"
  - "~status"
  - "~review"
  - "~scan"
  - "~conventions"
  - "~graph"
  - "~dashboard"
  - "~memory"
  - "~rlm"
  - "~validatekb"
---

# AgentFlow Skill

AgentFlow is a multi-CLI agent workflow system that keeps going until tasks are implemented and verified.

## Features

- **5-Level Routing (R0–R4)**: Every input is scored on 5 dimensions and routed proportionally — simple queries stay fast, complex tasks get full process, architecture-level tasks get dedicated review.
- **RLM Sub-Agent Orchestration**: 6 specialized roles (reviewer, synthesizer, kb_keeper, pkg_keeper, writer, architect) plus native CLI sub-agents, dispatched based on task complexity.
- **EHRB Safety Detection**: Three-layer safety (keyword scan, semantic analysis, tool output inspection) catches destructive operations before execution.
- **Three-Layer Memory**: L0 user memory, L1 project knowledge base, L2 session summaries — context survives across sessions.
- **Knowledge Graph Memory**: Graph-based project memory with progressive disclosure queries (AgentFlow unique).
- **Convention Extraction**: Automatic coding pattern discovery from your codebase (AgentFlow unique).
- **Architecture Scanning**: Proactive issue detection — large files, circular deps, missing tests (AgentFlow unique).

## Recommended MCP Servers

- **Context7 (Strongly recommended)**: for dependency / API / configuration documentation. Tip: append `use context7` in prompts, or provide a Context7 library ID (e.g. `/vercel/next.js`).
- **Playwright MCP (Recommended)**: for browser automation workflows.

## Quick Commands

| Command | Description |
|---------|-------------|
| `~init` | Initialize project knowledge base |
| `~auto` | Auto-execute with full workflow |
| `~plan` | Plan only, stop before development |
| `~exec` | Execute existing plan |
| `~status` | Show workflow status |
| `~review` | Code review |
| `~scan` | Architecture scan (AgentFlow unique) |
| `~conventions` | Extract/check coding conventions (AgentFlow unique) |
| `~graph` | Knowledge graph operations (AgentFlow unique) |
| `~dashboard` | Generate project dashboard (AgentFlow unique) |
| `~memory` | Manage memory layers |
| `~rlm` | Sub-agent management |
| `~validatekb` | Validate knowledge base consistency |

登录后下载/点赞/收藏 ❤ 13 | ★ 0

agentflow

目录结构

SKILL.md

举报内容

提示