jujutsu

The Jujutsu (jj) skill enables AI agents to utilize a Git-compatible version control system designed for high-concurrency development and streamlined operations. Unlike traditional systems that rely on staging areas, Jujutsu treats changes as immediate and mutable, allowing agents to manipulate repository history with greater flexibility. This skill provides a comprehensive set of operations for managing the complete lifecycle of a software project, from basic initialization to complex historical rewrites and remote synchronization.

• Perform repository operations such as initializing Jujutsu within existing Git projects or creating side-by-side colocated setups.
• Navigate commit history using logs, diffs, and change-id based switching, including time travel capabilities to traverse parent and child revisions.
• Execute advanced commit management including squashing changes into parents, splitting large changes into smaller logical units, and using absorb to automatically organize stacks of mutable revisions.
• Manage branching and remote interaction through bookmark creation, tracking, pushing to Git remotes, and fetching latest updates.
• Facilitate conflict resolution with interactive tools, perform rebasing of branches onto specific targets, and conduct regression testing via bisect.
• Enhance code quality by running configured formatters with fix, signing revisions cryptographically, and editing metadata without modifying underlying file content.

When using this skill, agents should prioritize the use of conventional commit messages to maintain clean project logs. Operations like undo and the operation log allow for safe experimentation, ensuring that agents can recover from complex rebasing or merging steps. Agents should be aware that Jujutsu changes are immediate; there is no staging step required before creating a new change or updating a description. This skill is ideal for development scenarios requiring rapid iteration, non-linear history management, and scenarios involving distributed remote collaboration. Inputs generally involve specific revision identifiers, commit messages, or target branches, while outputs reflect updated state in the repository or successful remote synchronization.