How does it detect 'two-lock' cascades?

It uses SQL windowing functions within DuckDB to identify consecutive reverts within a short timeframe, signaling that the AI is attempting the same rejected action repeatedly.

Does Amp-Skill require an external database setup?

No, it processes local Amp file-change history and indexes it into a local DuckDB instance for high-performance querying and analysis.

What is the primary purpose of Amp-Skill?

Amp-Skill is designed to analyze historical interaction data from Amp threads to identify patterns where an AI agent's actions were interrupted, reverted, or abandoned by the user.

Which file types are most sensitive to AI edits according to this skill?

Based on benchmark metrics, .org (Emacs org-mode) and .py files show higher rejection rates, suggesting these areas require more cautious AI behavior.

What is the Triadic Protocol integration?

It uses a GF(3) Trit system (Minus, Ergodic, Plus) to coordinate how the agent should react to rejections, such as validating the pattern or generating alternative solutions.

Amp Interruption Analyzer

Name: Amp Interruption Analyzer
Author: plurigrid

byplurigrid

•

分析与监控

Analyzes tool rejection patterns and agent interruption history to optimize AI behavior and reduce repetitive errors.

Amp-Skill provides deep insights into AI agent interactions by distilling usage patterns from Amp thread history. By leveraging DuckDB to analyze file-change data, it identifies critical interruption patterns such as two-lock cascades, abandoned threads, and reverted operations. This allows developers to pinpoint exactly where an agent’s suggestions are being rejected—particularly in sensitive file types like Org-mode or shell scripts—facilitating the implementation of smarter backoff strategies and more reliable autonomous workflows through the Triadic Protocol.

主要功能

01GF(3) Trit coordination layer for triadic protocol role assignment

02Interruption pattern detection using DuckDB SQL queries

038 GitHub stars

04Identification of two-lock cascades to prevent repetitive AI loops

05File-type sensitivity tracking for high-risk or personal documentation edits

06Historical analysis of tool call rejections and reverted operations

使用场景

01Debugging AI agent loops where tools are repeatedly rejected by the user

02Implementing automated backoff strategies for sensitive system file modifications

03Optimizing autonomous agent behavior based on historical abandonment rates

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi amp-skill

For use in Claude.ai and ChatGPT

Download Skill