grey-haven-context-management

📁 greyhaven-ai/claude-code-config 📅 8 days ago
2
总安装量
2
周安装量
#66036
全站排名
安装命令
npx skills add https://github.com/greyhaven-ai/claude-code-config --skill grey-haven-context-management

Agent 安装分布

opencode 2
amp 1
trae 1
kimi-cli 1
codex 1

Skill 文档

Context Management Skill

Multi-agent workflow orchestration with state persistence, handoff patterns, and resumable workflows.

Description

This skill enables sophisticated multi-agent workflows through context save/restore operations, supporting sequential handoffs, parallel execution, conditional routing, and long-running resumable workflows.

Token Efficiency Awareness

Multi-agent architectures have significant token costs. Understand the multipliers:

Architecture Token Multiplier Example (10K input)
Single agent 1x 10K tokens
Agent + tools 4x 40K tokens
Multi-agent 15x 150K tokens

Key insight: A 3-agent workflow costs ~15x more than a single agent. Design accordingly.

What’s Included

Examples (examples/)

  • Sequential handoff – Linear 4-phase workflow (Design → Implement → Test → Deploy)
  • Parallel execution – Full-stack feature with frontend + backend + tests (2.25x speedup)
  • Conditional routing – Dynamic agent selection based on results
  • Resumable workflows – Checkpoint-based long-running tasks

Reference Guides (reference/)

  • Complete reference – Save/restore operations, workflow patterns, performance optimization
  • Token efficiency – Token multipliers, attention budget, context optimization strategies
  • Troubleshooting guide – 6 issue categories with solutions (save failures, restore failures, performance, corruption)
  • Best practices – Context design, handoff patterns, error handling, security

Templates (templates/)

  • JSON Schema – Context structure validation with required/optional fields
  • Context templates – Ready-to-use structures for different workflow types

Checklists (checklists/)

  • 100-point verification – Comprehensive validation across 10 categories with weighted scoring

Workflow Patterns

  1. Sequential Handoff – Agent A → Context Save → Agent B → Context Save → Agent C
  2. Parallel Execution – Parent spawns 3 agents → Merge results → Integration
  3. Conditional Routing – Analyze → Route based on conditions → Targeted agent
  4. Resumable Long-Running – Checkpoints enable resume after interruption

Performance Metrics

  • Context Size: Target <100KB for 80% of workflows
  • Serialization: <200ms (P99)
  • Restore: <500ms (P99)
  • Speedup: 2-3x faster with parallel execution

Use This Skill When

  • Building multi-agent workflows
  • Need state persistence across sessions
  • Coordinating multiple agents on complex tasks
  • Implementing resumable long-running processes
  • Optimizing workflow execution time

Related Agents

  • context-manager – Automated context save/restore with validation
  • tdd-orchestrator – Multi-agent TDD workflows
  • incident-responder – Multi-phase incident response

Quick Start

# View workflow examples
cat examples/feature-development-handoff.md
cat examples/parallel-workflow-example.md

# Check reference guides
cat reference/INDEX.md
cat reference/workflow-best-practices.md

# Use context schema
cat templates/context-schema-template.json

# Validate context
cat checklists/context-verification-checklist.md

Security

  • ✅ Never save secrets in context (use secret references)
  • ✅ Implement access control for context files
  • ✅ Sanitize before saving (remove API keys, passwords)
  • ✅ Use relative file paths (not absolute)

Skill Version: 1.0 Last Updated: 2025-01-15