cognitive-architecture

AI Workflow Architecture — Agent Skills for Context Engineering

A comprehensive, open collection of Agent Skills combining context engineering knowledge with a standardized AI workflow architecture for building production-grade agent systems. Deployable to Claude Code, Gemini CLI, OpenCode, and Codex.

18 skills · 9 slash commands · L0–L3 workflow hierarchy · Claude → Gemini portation

Quick Start

Claude Code (Slash Commands)

# Install workflow commands locally
bash bin/install.sh --claude-local

# Then use:
/workflow:orchestrate "Build a REST API with authentication"
/workflow:progress
/workflow:port ./my-claude-skill

npm / npx

npx ai-workflow-skills --all --local

Plugin Marketplace (knowledge skills only)

/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering
/plugin install workflow-execution-layer@ai-workflow-architecture
/plugin install context-engineering-fundamentals@ai-workflow-architecture

---

What is Context Engineering?

Context engineering is the discipline of managing the language model's context window. Unlike prompt engineering, which focuses on crafting effective instructions, context engineering addresses the holistic curation of all information that enters the model's limited attention budget: system prompts, tool definitions, retrieved documents, message history, and tool outputs.

The fundamental challenge is that context windows are constrained not by raw token capacity but by attention mechanics. As context length increases, models exhibit predictable degradation patterns: the "lost-in-the-middle" phenomenon, U-shaped attention curves, and attention scarcity. Effective context engineering means finding the smallest possible set of high-signal tokens that maximize the likelihood of desired outcomes.

Recognition

This repository is cited in academic research as foundational work on static skill architecture:

"While static skills are well-recognized [Anthropic, 2025b; Muratcan Koylan, 2025], MCE is among the first to dynamically evolve them, bridging manual skill engineering and autonomous self-improvement."

— Meta Context Engineering via Agentic Skill Evolution, Peking University State Key Laboratory of General Artificial Intelligence (2026)

Skills Overview

Tier 1 — Workflow Execution (New in v2.0)

The operational backbone: an L0–L3 agent hierarchy implementing the Universal Agent Workflow standard.

Skill	Level	Description
pipeline-orchestrator	L0	Epic decomposition, kanban_board.md state management, delegation orchestration
story-executor	L1	Epic-to-story decomposition, L3 worker routing
task-reviewer	L2	Exclusive git commit authority; clean code + AC validation
task-executor	L3	Non-committing code generation; minimal-footprint implementation
task-rework	L3	Defect repair with rework-loop detection
test-executor	L3	Non-committing test suite creation; risk-based coverage

Tier 2 — Cross-Platform Migration (New in v2.0)

Skill	Description
cross-skill-porter	5-phase Claude Code → Gemini CLI portation pipeline. Non-destructive. Outputs TEST_RESULTS.md.
universal-agent-workflow	Binding standard: abstraction hierarchy, state tracking, Non-Commit Policy

Tier 3 — Context Engineering Knowledge Base

Foundational:

Skill	Description
context-fundamentals	Context window anatomy, attention mechanics, progressive disclosure
context-degradation	Lost-in-middle, context poisoning, distraction, and clash patterns
context-compression	Compression strategies, tokens-per-task optimization, probe-based evaluation

Architectural:

Skill	Description
multi-agent-patterns	Supervisor, swarm, and hierarchical multi-agent architectures
memory-systems	Temporal knowledge graphs, vector stores, file-system-as-memory
tool-design	Consolidation principle, MCP integration, tool naming conventions

Operational:

Skill	Description
context-optimization	KV-cache prefix caching, observation masking, context partitioning
evaluation	Multi-dimensional rubrics, LLM-as-judge patterns
advanced-evaluation	Pairwise comparison, position bias mitigation, production evaluation

Methodology: