skills.rest
kwcantrell
kwcantrellCOMMUNITY
Explore all skills by this author →
GitHub

context-evolve

Adjust context curation parameters and code to minimize token usage.

Updated Jul 5, 2026
Search Tags:#context management#runtime optimization#runtime evaluation#context drift reduction#token usage reduction
One-click install
npx skills add https://github.com/kwcantrell/rusty-agent --skill context-evolve
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-evolve
Source: https://github.com/kwcantrell/rusty-agent/tree/main/.agents/skills/context-evolve
Command: npx skills add https://github.com/kwcantrell/rusty-agent --skill context-evolve

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires agent-cli, llama-server, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill helps optimize the context management subsystem of the runtime, improving task-solving efficiency by minimizing context drift and reducing token usage without compromising correctness.

Core Features & Use Cases

  • Context Optimization: Iteratively adjusts curation parameters and code to improve long-term memory and in-window curation.
  • Task Evaluation: Evaluates changes against live models on frozen tasks to ensure correctness and efficiency.
  • Use Case: For a developer looking to optimize the long-term memory and in-window curation of their runtime's context management system, this skill can automatically test changes and ensure that optimizations do not compromise correctness while reducing token usage.

Quick Start

To prepare a new task, run the 'prepare' command. To optimize the context manager, execute the 'train' command.

Frequently Asked Questions about context-evolve

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

FAQPage Schema
How do I optimize context management to reduce token usage in my runtime?▼

Runtime context management optimization is achieved by iteratively adjusting curation parameters and code to minimize token usage. This process evaluates changes on frozen tasks against live models to ensure correctness is not compromised while reducing context drift.

What causes context drift and how does evaluating frozen tasks help fix it?▼

Context drift occurs when runtime curation parameters degrade task-solving efficiency over time. Evaluating changes on frozen tasks against live models measures this drift, ensuring that long-term memory and in-window curation adjustments improve efficiency without breaking correctness.

How do I test context manager changes without compromising task correctness?▼

Testing context manager changes without compromising correctness requires evaluating adjustments on frozen tasks against live models. This validation ensures that any code or parameter modifications made to reduce token usage maintain the expected task-solving outcomes.

Do I need agent-cli and llama-server to run context optimization tasks?▼

Yes, agent-cli and llama-server are required dependencies to run context optimization tasks. These tools facilitate the execution of the prepare and train commands, which set up frozen tasks and apply curation parameter adjustments against live models.

What is the best way to enhance long-term memory and in-window curation for a runtime?▼

The best way to enhance long-term memory and in-window curation is through iterative context optimization. By running the train command, the system automatically adjusts curation parameters and code, validating improvements against frozen tasks to maximize runtime efficiency.

When should I not use automated context parameter adjustments for runtime optimization?▼

Automated context parameter adjustments should be avoided when you lack frozen tasks for validation. Without evaluating changes against live models on frozen tasks, automated adjustments risk increasing context drift and compromising task correctness.

Related Skills

Explore More Software Engineering Skills→
servitolaservitola
12

llama-cpp

Guide GGUF quant selection and launch llama.cpp server or CLI inference.

Community
Advanced
Sir-chawakornSir-chawakorn
33

memory-system

Manage AI context and memory persistence using structured markdown files.

Community
Intermediate
haidonglethqbhaidonglethqb

context-engineering

Optimize LLM context windows with deterministic token budgeting and probe-based evaluation.

Community
Advanced
Shadojin94Shadojin94

performance-optimizer

Reduce OpenClaw agent token usage with caching and context guards.

Community
Advanced