project-evaluation

bjpl 的头像作者 bjpl

"使用蜂群协调、GOAP规划、神经分析和AgentDB内存进行全面项目状态评估。当评估架构健康状况、规划重构或生成状态报告时使用。"

📖 中文导读:"使用蜂群协调、GOAP规划、神经分析和AgentDB内存进行全面项目状态评估。当评估架构健康状况、规划重构或生成状态报告时使用。" 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: "Project Evaluation" description: "Comprehensive project status evaluation using hive-mind coordination, GOAP planning, neural analysis, and AgentDB memory. Use when assessing architecture health, planning refactoring, or generating status reports."

Project Evaluation Skill

What This Skill Does

Orchestrates comprehensive project evaluation using all Claude Flow systems:

  • Hive-Mind: Collective intelligence coordination
  • AgentDB: Persistent memory across sessions
  • Neural Training: Pattern learning from evaluations
  • GOAP Planning: Action planning for improvements
  • Skill Creation: Document learnings as reusable skills

Prerequisites

  • Claude Flow v2.0+ (npx claude-flow@alpha)
  • Initialized hive-mind (npx claude-flow hive-mind init)
  • Project with prior session memory

Quick Start

# 1. Retrieve prior session state
npx claude-flow memory retrieve --namespace hive-mind --key "session/*"

# 2. Initialize evaluation swarm
npx claude-flow swarm init --topology hierarchical --agents 5

# 3. Spawn evaluation agents
# Use Claude Code Task tool:
Task("Architecture Agent", "Evaluate architecture health...", "system-architect")
Task("GOAP Agent", "Generate improvement plan...", "code-goal-planner")

Evaluation Framework

Phase 1: Memory Retrieval

// Retrieve prior session state
const session = await memory.retrieve('session/*/completed', 'hive-mind');
const worldState = await memory.retrieve('goap/world-state/final', 'goap');

Phase 2: Agent Spawning

// Spawn evaluation agents via Task tool
[Parallel]:
  Task("system-architect", "Evaluate architecture against assessment...")
  Task("code-goal-planner", "Generate GOAP plan for Grade A...")
  Task("tester", "Analyze test coverage and quality...")

Phase 3: Neural Analysis

// Train on evaluation patterns
await neuralTrain({
  pattern_type: "coordination",
  training_data: { metrics: ["architecture", "testing", "performance"] }
});

Phase 4: Results Storage

// Store in AgentDB for persistence
await memory.store('evaluation/architecture-grade', results, 'agentdb');
await memory.store('evaluation/goap-plan', plan, 'goap');

Evaluation Metrics

Category Metrics
Architecture Grade, critical issues, domain separation
Testing File count, coverage %, pass rate
Performance Bundle size, build time, store LOC
Tech Debt Deprecated code, uncommitted changes

Output Format

{
  "evaluation": {
    "previousGrade": "B-",
    "currentGrade": "B+",
    "criticalIssuesResolved": 3,
    "criticalIssuesRemaining": 0
  },
  "goapPlan": {
    "totalCost": 13,
    "actions": ["commit", "test", "delete", "optimize"],
    "successProbability": "85%"
  },
  "metrics": {
    "storeLOC": 3678,
    "testFiles": 67,
    "uncommittedFiles": 19
  }
}

Integration with Other Skills

  • store-migration-workflow - For refactoring execution
  • hive-mind-advanced - For collective coordination
  • agentdb-memory-patterns - For persistence
  • goap-planning - For action sequencing

Best Practices

  1. Always retrieve prior session state before evaluation
  2. Store all findings in AgentDB for cross-session persistence
  3. Train neural patterns on successful evaluations
  4. Generate GOAP plans for actionable next steps
  5. Create skills from recurring evaluation patterns

Created: 2025-12-03 Version: 1.0.0 Category: Project Management