skills.rest
helderberto
helderbertoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

context-engineer

Manage and optimize agent context loading with rules-first and just-in-time protocols.

12|1|Updated Feb 13, 2026
Search Tags:#context#agent#conventions#session-recovery#just-in-time#load-rules
One-click install
npx skills add https://github.com/helderberto/agent-skills --skill context-engineer-helderberto
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-engineer
Source: https://github.com/helderberto/agent-skills/tree/main/skills/context-engineer
Command: npx skills add https://github.com/helderberto/agent-skills --skill context-engineer-helderberto

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Context is the agent's working memory. Without disciplined management, outputs can hallucinate, violate conventions, and drift across tasks. This skill provides a principled approach to decide what to load, in what order, and what to drop as context evolves.

Core Features & Use Cases

  • Rules-first loading: always load project-wide rules, standards, and reference policies.
  • Just-in-time loading: pull files or snippets only when needed for a task.
  • Recovery & session switching: recover from degraded outputs and reset when pivoting tasks.
  • Guided session protocol: defines a structured start sequence for non-trivial work.

Quick Start

To begin, state the goal, ensure rules are loaded, and load the most relevant files for the task.

Frequently Asked Questions about context-engineer

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

FAQPage Schema
Why does my agent hallucinate or violate project conventions during complex tasks?▼

Agent hallucinations occur when working memory lacks structure. Applying a rules-first load order and just-in-time content loading ensures standards are always present, preventing drift during complex tasks.

How do I manage agent context when the context window is near its limit?▼

To manage context limits, apply just-in-time loading to pull files only when needed. This optimizes what loads into the agent's context, preventing overload and ensuring stable reasoning.

What is the best way to recover an agent session after switching tasks?▼

The best way to recover an agent session after switching tasks is to execute a structured session-start protocol. This resets the agent context, applies error recovery, and reloads rules to prevent output drift.

Do I need a structured session protocol for every agent interaction?▼

A structured session protocol is not required for every interaction. Apply it during non-trivial work, complex tasks, session switches, or when the context window nears its limit to ensure stable reasoning.

How to start a complex agent task while preserving project conventions?▼

To start a complex task while preserving conventions, state the goal, ensure project-wide rules are loaded first, and load the most relevant files. This guided protocol enforces a rules-first load order.

Related Skills

Explore More Software Engineering Skills→
GoBeromsuGoBeromsu
2

context-engineering

Structure and persist project-wide context rules across agent sessions.

Community
Advanced
cassioalexandrecassioalexandre
2

llm-prompt-optimizer

Rewrite prompts using the RSCIT framework to enforce structured outputs and reduce hallucinations.

Community
Intermediate
kivo360kivo360
1

moai-cc-memory-arch

Manage Claude Code session memory with just-in-time retrieval and context summaries.

Community
Advanced
NackalalalongNackalalalong

Agent Tool Calling

Automate a self-contained agent tool-calling loop with memory and streaming.

Community
Advanced