Skip to content
SkillModel-led1,448 tokens in use

PR Improver

Runs a review-and-fix loop over the current branch's changes until a PR review reports no critical or major findings, scoped to the touched directories.

8,319 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/code-improver/skills/pr-improver
/plugin marketplace add trailofbits/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing PR Improver costs

108tokens idle
1,448tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldeclared and detectedwrites filesdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Excellent
  • Frontmatter8/15
  • Structure25/25
  • Examples12/15
  • Output15/15
  • Size15/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity100
  • Maintenance52
  • Community64
  • Maturity32
  • Momentum40

PR Improver cleans up a branch before you open or update a pull request. It derives the change surface from the git diff against the base branch, widens each changed file to its directory, and runs a Workflow loop where a PR reviewer and a fixer subagent alternate until the review is clean.

The default reviewer is the review-pr skill from pr-review-toolkit, and you can name another skill or agent. All edits stay in the working tree and nothing is committed.

What You Get

  • A scope derived from the branch diff and deduplicated into directory globs
  • Up to five rounds of review and fix, or the cap you set
  • A ledger and metrics file for the run, plus a note on any warning such as a repository being initialized
  • A labelled outcome: converged, capped and not converged, escalation or halted

Common Use Cases

  • Fixing review findings on a feature branch before opening a pull request
  • Running review-and-fix on local changes against main
  • Using a different PR review agent than the default one
  • Continuing after an escalation with your ruling passed as a decision

What Makes It Unique

The scope comes from your branch and is enforced mechanically after each round, so the fixer cannot wander into unrelated directories. The skill refuses to imitate the loop by hand when Workflow is unavailable and tells you which plugin provides a missing reviewer.

Under The Hood

The skill runs git commands to find the repository root, base branch and changed files, then starts a Workflow script from the plugin. Inside the workflow a reviewer and a fixer subagent read and edit files in the derived directories, and a ledger and metrics are written to disk. It checks scope with git after each fix round. It does not commit, push or open a pull request, and it does not fetch web pages or call outside APIs by itself.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers