ax-interview

Analyze agent-tool interaction traces using Gricean maxims to detect friction and generate prioritized fixes.

1|Updated Mar 9, 2026
One-click install
npx skills add https://github.com/Peleke/brunnr --skill ax-interview
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ax-interview
Source: https://github.com/Peleke/brunnr/tree/main/skills/ax-interview
Command: npx skills add https://github.com/Peleke/brunnr --skill ax-interview

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Run a structured post-task AX Interview that surfaces friction invisible to evaluations by applying Gricean maxims to agent-tool interactions and analyzing behavioral signals against traces.

Core Features & Use Cases

  • Phase-based trace analysis including maxim violations, implicature detection, dialog probes, cross-reference checks, and Pragmatic Coherence Score reporting.
  • Useful after multi-step agent tasks to reveal hidden performance issues and generate prioritized fixes.

Quick Start

After a task session, run the AX Interview to generate a Pragmatic Coherence Score and a ranked list of fixes.

Frequently Asked Questions about ax-interview

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

FAQPage Schema
How do I analyze agent-tool interaction traces to find hidden friction?

Evaluate tool friction by running a structured post-task AX Interview that applies Gricean maxims to dialog and traces, yielding a Pragmatic Coherence Score and prioritized fixes.

What is a Pragmatic Coherence Score in agent evaluation?

A Pragmatic Coherence Score is a metric generated by analyzing behavioral signals against trace data to measure pragmatic coherence and reveal hidden performance issues in agent-tool interactions.

Can I evaluate agent tool friction with dialog only if trace data is unavailable?

Yes, you can run the AX Interview in dialog-only mode when traces are unavailable, applying Gricean maxims to detect friction and generate a Pragmatic Coherence Score from the completed session.

When should I run a Gricean maxim analysis on agent traces?

Run Gricean maxim analysis after a completed multi-step agent task session to evaluate agent-tool interactions and generate prioritized fixes for hidden performance issues.

What inputs do I need to perform a trace analysis for agent-tool interactions?

Required inputs include a completed task session, trace data, and the tools under evaluation; if these are missing, the session will prompt for the necessary data.