skills.rest
sourcegraph
sourcegraphOFFICIAL
Explore all skills by this author →
Website𝕏@sourcegraphGitHub

quick-rerun

Execute a single benchmark task locally with the Harbor runner.

32|4|Updated Feb 1, 2026
Search Tags:#testing#debugging#verification#benchmark#ci#local run
One-click install
npx skills add https://github.com/sourcegraph/CodeScaleBench --skill quick-rerun
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: quick-rerun
Source: https://github.com/sourcegraph/CodeScaleBench/tree/main/skills/quick-rerun
Command: npx skills add https://github.com/sourcegraph/CodeScaleBench --skill quick-rerun

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill allows developers to quickly re-run a specific benchmark task locally to verify that a code fix has resolved an issue, streamlining the debugging and verification process.

Core Features & Use Cases

  • Local Task Execution: Run individual benchmark tasks without needing the full Harbor environment.
  • Fast Verification: Utilizes claude-haiku for rapid iteration and verification of fixes.
  • Use Case: After fixing a bug in a benchmark task, use this Skill to run just that task to confirm the fix before committing or proceeding with a full benchmark run.

Quick Start

Use the quick-rerun skill to verify the fix for task 'sgt-005'.

Frequently Asked Questions about quick-rerun

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

FAQPage Schema
How do I run a single benchmark task locally to verify a code fix?▼

To run a single benchmark task locally for fix verification, execute the task using the Harbor runner. This automates task path resolution, environment setup, and result checking via result.json analysis.

Can I use different models for local benchmark task verification?▼

Local benchmark task verification supports configurable models like Claude Haiku for rapid iteration or Opus for full execution. You can also choose between baseline execution without MCP and full MCP execution modes.

What is the best way to debug a specific benchmark task without running the full environment?▼

The best way to debug a specific benchmark task without the full environment is to execute individual tasks locally. This isolates the task path and automates environment setup to verify code fixes quickly.

Do I need the full Harbor environment to execute benchmark tasks for debugging?▼

You do not need the full Harbor environment to execute benchmark tasks for debugging. This skill automates local task execution independently to verify fixes before committing or running a full benchmark suite.

How does result checking work after executing a local benchmark task?▼

Result checking after executing a local benchmark task works by analyzing the result.json file. The automation framework handles this analysis automatically to confirm whether your code fix resolved the issue.

Why use quick-rerun instead of running a full benchmark suite after fixing a bug?▼

Use quick-rerun instead of a full benchmark suite to rapidly verify a single bug fix locally. It leverages Claude Haiku for fast iteration and skips the overhead of a complete benchmark run.

Related Skills

Explore More Software Engineering Skills→
qiushiyanqiushiyan

harbor-task-creator

Automate creation and management of AI agent evaluation tasks in Harbor.

Community
Advanced
e-burgose-burgos
2

monitor-ci

Monitor Nx Cloud CI pipelines and trigger self-healing workflows on failures.

Community
Advanced
onchainengineeronchainengineer

tbench

Automate Unix agent benchmarking with Terminal-Bench 2.0 on Docker or Daytona.

Community
Advanced
sourcegraphsourcegraph
32

scaffold-task

Generate Harbor-compatible CodeScaleBench task files and suite configurations.

Official
Advanced