0.71.0

underpig1/octos0.71.0Feb 17, 2026by takahirom

AI Summary

This release introduces new command-line capabilities for executing tasks and enhances configuration management by adding tags to result files, alongside establishing benchmarking infrastructure and workflow improvements.

Key Highlights

Added 'run task' subcommand for single-goal execution
Added tags to result.yml for better scenario management
Added AndroidWorld benchmark CI workflow
Prevented Legal information scenario from opening Third-party licenses

New Features

'run task' subcommand
Tags in result.yml
AndroidWorld benchmark CI
Prevention of opening Third-party licenses in Legal scenario

Full Release Notes

* We've added the `arbigent run task "goal"` command so you can easily try an Arbigent task. This lets us measure Arbigent’s performance with benchmarks.
* We've added tags to result.yml so you can manage scenarios with other tools.

## What's Changed
* Add 'run task' subcommand for single-goal execution by @takahirom in https://github.com/takahirom/arbigent/pull/366
* Add AndroidWorld benchmark CI workflow by @takahirom in https://github.com/takahirom/arbigent/pull/367
* Prevent Legal information scenario from opening Third-party licenses by @takahirom in https://github.com/takahirom/arbigent/pull/370
* Add tags to ArbigentScenarioResult in result.yml by @takahirom in https://github.com/takahirom/arbigent/pull/371


**Full Changelog**: https://github.com/takahirom/arbigent/compare/0.70.0...0.71.0