context-guardian

Audit CLAUDE.md and prompts to prune context bloat and output a structured Context Audit Report.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/nkootstra/skills --skill context-guardian
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: context-guardian
Source: https://github.com/nkootstra/skills/tree/main/skills/context-guardian
Command: npx skills add https://github.com/nkootstra/skills --skill context-guardian

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Context bloat is the number one reason agents underperform: unrelated rules, stale memory, and noisy prompts pollute task execution. Context Guardian helps developers audit CLAUDE.md and prompts to prune unnecessary information while preserving essential context.

Core Features & Use Cases

  • Context audits: prune CLAUDE.md and prompts to essential rules relevant to the current task.
  • Phase separation guidance: split vague tasks into research (discovery) and implementation (fresh context) phases.
  • Session drift mitigation: provide compaction and re-reading rules to prevent memory loss during long sessions.

Quick Start

Take your current CLAUDE.md or prompt, identify bloat, and output a concise, phase-separated action plan.

Frequently Asked Questions about context-guardian

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

FAQPage Schema
How do I reduce context bloat in CLAUDE.md to improve agent performance?

Reduce context bloat by auditing CLAUDE.md to prune unrelated rules and stale memory. This preserves essential context, prevents session drift, and maintains peak agent performance during task execution.

What is the best way to prevent session drift during long agent conversations?

Prevent session drift by applying compaction and re-reading rules to mitigate memory loss. Splitting vague tasks into research and implementation phases ensures fresh context for execution.

How do I audit prompts for noise and unnecessary information?

Audit prompts for noise by applying a prune-focused workflow that identifies and removes unrelated information. This outputs a structured Context Audit Report categorizing kept, moved, and removed items.

Why does my agent underperform when handling vague software engineering tasks?

Agents underperform with vague tasks due to context pollution from noisy prompts and a lack of phase separation. Decomposing tasks into distinct research and implementation phases restores execution accuracy.

Can I split agent tasks into research and implementation phases to save context?

Yes, you can split tasks into research and implementation phases. This phase separation guidance ensures discovery happens first, allowing implementation to execute with fresh, lean context.

What does a Context Audit Report include when pruning prompt-engineering rules?

A Context Audit Report includes structured lists of kept, moved, and removed items. It categorizes essential rules while identifying bloat for pruning to minimize session drift and optimize execution.