context-audit

Audit default-loaded context files and emit per-file KEEP/TRIM/DEMOTE/CUT verdicts.

28|3|Updated Mar 27, 2026
One-click install
npx skills add https://github.com/mifunedev/openharness --skill context-audit-mifunedev
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: context-audit
Source: https://github.com/mifunedev/openharness/tree/main/.claude/skills/context-audit
Command: npx skills add https://github.com/mifunedev/openharness --skill context-audit-mifunedev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps teams automatically quantify how much context is loaded per file and converts it into actionable KEEP/TRIM/DEMOTE/CUT verdicts to help maintain safe, efficient agent behavior.

Core Features & Use Cases

  • Score the default-loaded context across four deterministic dimensions (footprint, load-bearing, integrity, redundancy) and emit per-file verdicts.
  • Optional Tier-2 ablation harness to verify proposed context cuts by removing a target file, running a fixed probe suite, and measuring behavior degradation.
  • Generate a machine-readable and human-friendly summary for governance, memory logging, and handoff to downstream nodes.

Quick Start

Run the context-audit skill against the harness to generate per-file verdict report and, if desired, baseline and ablation probe results.

Frequently Asked Questions about context-audit

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

FAQPage Schema
How do I audit context budgets and reduce loaded token bloat?

You can audit context budgets by scoring loaded files across four dimensions—footprint, load-bearing, integrity, and redundancy—to generate actionable KEEP, TRIM, DEMOTE, or CUT verdicts per file.

What does context ablation mean for agent memory governance?

Context ablation is a verification process that removes a target file, runs a fixed probe suite, and measures behavior degradation to validate proposed context cuts for memory governance.

How do I check if a specific memory file is safe to delete?

To check deletion safety, run an optional Tier-2 ablation harness that removes the target file, executes a fixed probe suite, and measures behavior degradation against a memory-durable baseline.

How are context files scored for load-bearing and redundancy?

Context files are scored across four deterministic dimensions: Footprint, Load-bearing, Integrity, and Redundancy. These dimensions are aggregated into a total score to determine KEEP or CUT verdicts.

Can I apply context auditing to standard loaded sets like CLAUDE.md?

Yes, context auditing applies directly to the harness standard loaded set, including CLAUDE.md, context directories, and memory files, generating per-file verdicts and a tiered summary report.