Skip to content

GIT Committer Skill: High Usage #2

Description

@vigilantloki

The /git-commiter enables LLM agents to commit with a specified ruleset and formatting. However, it was observed that the skill consumes more tokens than expected - which should not be the case. Probable cause for the high consumption:

Getting List of Changes

The skill accesses all the files to be staged for the current commit to understand what was actually changed. This is a "guard rail" since there might be some changes between each session and user's intervention which will be tracked during the commit, else there might be an issue in reference from what was the commit message vs what was done.

2. Run `git diff --stat` (and `git diff <file>` for each file) to understand what actually changed.

Commit Topology

The skill tries to group related changes in one commit rather than creating a huge commit of all the files. The logical grouping allows sequential tracking of the developmental changes in the directory and exact recovery mechanism.

3. Decide the **commit topology**: one focused commit, or one commit per logically independent file.

Metadata

Metadata

Assignees

No one assigned

    Labels

    extensions/skillsAn issue/PR is related to fixing a "skills" from the extension's library.issue/t/performanceImprovement of performance of code time/memory benchmarking.

    Type

    Fields

    No fields configured for P0.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions