ScavioScavio
Workflow

Agent Skill Trim Workflow

Quarterly: count skill description tokens, drop never-invoked, drop duplicates, replace narrow web skills with Scavio MCP, re-measure.

Start FreeAPI Docs

Overview

Quarterly trim discipline for agent runtimes that suffer skill bloat (Claude Code, Hermes, OpenClaw). Cuts per-message input tokens; the trim is the work, not the tools.

Trigger

Quarterly (manual)

Schedule

Quarterly

Workflow Steps

1

Count description tokens per skill

Top 10 by weight = scrutinize first.

2

Drop skills not invoked in 2 weeks

Honest log = honest list.

3

Drop duplicate skills

Multiple 'search' / 'fetch URL' variants → keep one.

4

Replace 5-8 narrow web/scrape skills with one Scavio MCP

Reduces per-message input by ~5-8K tokens.

5

Re-measure input token cost per message

Before vs after; track weekly.

6

Re-audit quarterly

Drift back is real.

Python Implementation

Python
# Process workflow. Actionable parts via CLI:
# claude mcp list
# claude mcp remove <unused>
# claude mcp add scavio https://mcp.scavio.dev/mcp --header 'x-api-key: $SCAVIO_API_KEY'

JavaScript Implementation

JavaScript
// CLI discipline.

Platforms Used

Google

Web search with knowledge graph, PAA, and AI overviews

Frequently Asked Questions

Quarterly trim discipline for agent runtimes that suffer skill bloat (Claude Code, Hermes, OpenClaw). Cuts per-message input tokens; the trim is the work, not the tools.

This workflow uses a quarterly (manual). Quarterly.

This workflow uses the following Scavio platforms: google. Each platform is called via the same unified API endpoint.

Yes. Scavio's free tier includes 500 credits per month with no credit card required. That is enough to test and validate this workflow before scaling it.

Agent Skill Trim Workflow

Quarterly: count skill description tokens, drop never-invoked, drop duplicates, replace narrow web skills with Scavio MCP, re-measure.

Get Started FreeRead the Docs