Do I need DuckDB to use this skill?

Yes, DuckDB is used as the underlying engine to perform complex SQL queries on your local JSON and JSONL datasets efficiently.

Can I see which tools are used most frequently?

Absolutely. The skill provides specific SQL queries to count invocations for core tools like Bash, edit_file, and various MCP servers.

What local data does this skill analyze?

It mines JSON thread data in ~/.local/share/amp/threads/, history logs in ~/.claude/history.jsonl, and project files to provide a unified view of your AI activity.

How does it estimate LLM costs?

The skill includes a pricing mapping for Claude 4.5 and Gemini 3 series, applying cost profiles based on whether the agent is in smart, rush, or large mode.

What are visibility levels in Amp threads?

Threads can be classified as private, workspace, group, public, or unlisted, all of which can be audited using the visibility analysis features.

Amp API Awareness & Analytics

Name: Amp API Awareness & Analytics
Author: plurigrid

byplurigrid

•

분석 및 모니터링

Analyzes local Amp thread data and API patterns using DuckDB to uncover usage statistics and hidden system schemas.

Amp API Awareness provides a powerful diagnostic and analytical layer for the Amp environment by mining local thread storage and session history. By leveraging DuckDB for high-performance JSON querying, it extracts undocumented API patterns, tool invocation frequencies, and MCP (Model Context Protocol) server interactions. This skill is essential for developers and teams who need to audit AI usage, estimate LLM API costs across different models like Claude 4.5 and Gemini 3, and understand the underlying data structures of their agentic workflows.

주요 기능

01Comprehensive tool usage tracking for Bash, file editing, and custom skills

02Detection and mapping of Model Context Protocol (MCP) server invocations

03Workspace discovery for identifying user IDs and thread visibility levels

04Automated DuckDB schema generation for local thread and history files

05Granular LLM cost estimation based on model-specific pricing (Opus, Sonnet, Haiku)

068 GitHub stars

사용 사례

01Analyzing agent performance by mining tool success rates and message patterns

02Auditing team-wide AI resource consumption and per-thread API costs

03Reverse-engineering undocumented Amp endpoints for workflow automation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi amp-api-awareness

For use in Claude.ai and ChatGPT

Download Skill