code-just-fix

Runs the project just check suite and fixes every failure at the root cause.

2|Updated Apr 2, 2026
One-click install
npx skills add https://github.com/nseng-ai/ns --skill code-just-fix-nseng-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-just-fix
Source: https://github.com/nseng-ai/ns/tree/main/skills/internal/code/code-just-fix
Command: npx skills add https://github.com/nseng-ai/ns --skill code-just-fix-nseng-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Failing CI-style checks (formatting, lint, type errors, broken tests) require tedious manual triage across many recipes. This Skill automates the loop of running the project's just suite, diagnosing each failure, and fixing the underlying code until everything passes. ## Core Features & Use Cases - Automated failure triage: Runs just, categorizes failures by recipe into formatting, lint, type/compile, and test failures, then applies the appropriate fix strategy for each. - Root-cause fixing: Fixes production code or tests based on where the real bug is, never deleting, skipping, or weakening tests and never suppressing errors with ignore comments. - Iteration limits and stuck detection: Stops after 10 iterations or when the same error repeats 3 times, then reports a structured stuck summary with attempted fixes and next steps. - Use Case: After a large refactor, invoke the Skill to run just, fix the two formatting issues and one type error it finds, and confirm all recipes pass in three iterations. ## Quick Start Ask the agent to run the project just suite and fix every failing check until all recipes pass.

Frequently Asked Questions about code-just-fix

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

FAQPage Schema
How do I automatically fix failing just recipes in my project?

Invoke the Skill to run `just`, which categorizes failures by recipe into formatting, lint, type, and test issues. It applies the appropriate fix for each category and re-runs the suite until all checks pass or an iteration limit is reached.

How to fix failing tests without deleting or skipping them?

Read both the failing test and the code under test to determine where the real bug is, then fix the production code or the test accordingly. The Skill explicitly forbids deleting, skipping, or weakening tests and suppressing errors with ignore comments.

Does this work with planning or read-only agent modes?

Yes. In planning mode it runs `just`, reads the relevant source and test files, and presents a structured plan listing each failure with file, line, root cause, and proposed fix without editing anything. The user approves before changes are applied.

What happens when a check keeps failing after multiple fix attempts?

The Skill stops after 10 iterations or immediately if the same error appears in 3 consecutive iterations. It then reports a structured stuck summary including the failing recipe, exact error, attempted fixes, and suggested next steps.

What tools are required to run the just fix workflow?

The project must use just as its task runner with a justfile defining the check recipes. The Skill uses Bash to run just and uv commands, plus file read, edit, glob, and grep capabilities to diagnose and fix failures.