SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI codingagent-operator-score

agent-operator-score

Local-first CLI (`aos`) that scores how well you operate Claude Code, Codex, and Grok CLI from your own session transcripts, plus a controlled-run assessment suite with a hidden verifier. No model calls in review mode; nothing leaves your disk.

Open website

About this tool

Purpose

agent-operator-score is a local-first command-line tool, mainly used to evaluate a developer's ability to use command-line AI coding assistants. According to available information, its evaluation targets include CLI-form AI coding tools such as Claude Code, Codex, and Grok CLI.

Main Capabilities

  1. Session transcript scoring: The tool reads the session transcripts already present on the user's local machine and uses them to score the developer's operating proficiency.
  2. Controlled-run evaluation suite: In addition to direct scoring, it also provides a controlled-run evaluation suite, which is equipped with a hidden verifier.

How to Use

It is used through a command-line interface, with the command aos. The tool is in CLI form and obtains data from local session transcripts.

Privacy and Runtime Characteristics

  • Local-first: In review mode, no model calls are made, and data does not leave the user's local disk.
  • The tool itself does not include any explanation regarding remote data upload behavior; the available information only emphasizes the local characteristics in review mode.

Applicable Scenarios

Suitable for developers who frequently use command-line AI coding assistants such as Claude Code, Codex, and Grok CLI, to quantify the effectiveness of their collaboration with these tools and to conduct stricter examination with the help of the controlled evaluation suite.

Limitations and Information Not Yet Provided

  • The specific scoring dimensions, scoring algorithm, and meaning of scores are not provided in the available information.
  • The specific usage steps of the controlled-run evaluation suite and the decision logic of the hidden verifier are not provided in the available information.
  • Supported platforms, installation methods, dependency environment, and license agreement are not provided in the available information.
  • Whether it is paid and its pricing plan are not provided in the available information.
  • Whether CLI tools other than Claude Code, Codex, and Grok CLI are supported is not provided in the available information.

Related tools

Browse category