skills.rest
Handy369
Handy369COMMUNITY
Explore all skills by this author →
GitHub

context-engineering

Diagnose and rebuild layered agent context from rules, specs, source, and history.

Updated Apr 27, 2026
Search Tags:#hallucination reduction#conversation management#task switching#agent context#rules files#project specifications
One-click install
npx skills add https://github.com/Handy369/passto-pi-frame --skill context-engineering-handy369
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-engineering
Source: https://github.com/Handy369/passto-pi-frame/tree/main/skills/context-engineering
Command: npx skills add https://github.com/Handy369/passto-pi-frame --skill context-engineering-handy369

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill fixes declining agent output quality by diagnosing whether the model is suffering from context starvation, flooding, staleness, or missing rules/config, then rebuilding the right layered context.

Core Features & Use Cases

  • Context triage and diagnosis: determines whether the current problem is starvation, flooding, stale context, task switching, or rules/config gaps.
  • Layered context assembly: selects and orders rules files, relevant specs/architecture docs, task-specific source files, error/test output, and compacted conversation history.
  • Hallucination and drift reduction: lowers noise, resolves conflicts explicitly, and refreshes context when switching projects or sub-systems.
  • When to use: starting a new session/project, detecting hallucinations or convention violations, and preparing the agent to work across code areas.

Quick Start

Ask the agent to reorganize your project context for the current task by diagnosing the context issue first, then loading the appropriate rules, specs, source files, and the most relevant history.

Frequently Asked Questions about context-engineering

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
Why does my AI agent hallucinate and ignore project rules during code generation?▼

Agent hallucination often stems from poor context setup like starvation, flooding, or stale information. Diagnosing the specific context issue and rebuilding a layered context with explicit rules and relevant source files reduces hallucination and drift.

How do I rebuild agent context when switching between different code areas?▼

To rebuild agent context for task switching, load layered rules files, relevant specs and architecture docs, task-specific source files, and compacted conversation history. This refreshes the agent's working state and resolves conflicts explicitly.

What is the best way to organize project specifications and rules files for AI agents?▼

Organizing project specifications requires layered context assembly that selects and orders rules files, architecture docs, and source files. This structure lowers noise and ensures the agent loads appropriate conventions for the current task.

How can I fix declining AI output quality in long coding sessions?▼

Declining AI output quality is fixed by diagnosing whether the model suffers from context starvation or flooding, then rebuilding the right layered context with compacted conversation history and relevant source files.

When do I need to reorganize my AI agent's conversation history and context?▼

You need to reorganize agent context when starting new sessions, detecting convention violations, experiencing output degradation, or preparing the agent to work across different code areas or sub-systems.

Can I use context engineering without external dependencies or components?▼

Yes, rebuilding agent context requires no external dependencies. It works by applying layered loading of existing project rules, specs, source files, and compacted conversation history with explicit conflict and gap handling.

Related Skills

Explore More Software Engineering Skills→
hgflimahgflima
1

context-engineering

Curate structured context from rules files, specs, and source files for AI agents.

Community
Advanced
hidonguyenhidonguyen

ck:test

Execute unit, integration, e2e, and UI tests across JavaScript/TypeScript, Python, Go, Rust, and Flutter projects.

Community
Advanced
pascallammerspascallammers
1

context-optimizer

Summarize conversation history and compact tool results to prevent context overflow.

Community
Advanced
NguyenSiTrungNguyenSiTrung
24

conductor

Load project context files and manage structured tracks and specs.

Community
Advanced