Skip to content
SkillModel-led1,386 tokens in use

Code Improver

Runs an autonomous review-and-fix loop over any code target with a reviewer you name, a findings ledger, escalation and a scope guard.

8,319 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/code-improver/skills/code-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 Code Improver costs

114tokens idle
1,386tokens 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

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
  • Frontmatter12/15
  • Structure23/25
  • 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

Code Improver improves a skill, plugin, module or directory by looping a reviewer and a fixer subagent until a review reports zero critical or major findings. You name the reviewer, any installed skill or agent, and the glob scope the loop may touch.

The loop runs as a background Workflow with a cross-round ledger, an escalation when fixes stop converging, and a mechanical scope check after every fix round. It never commits, and it ends by stripping its own narration comments.

What You Get

  • A loop that reviews, fixes and re-reviews up to a round cap, five by default
  • A findings ledger on disk that carries over between runs
  • Clear outcome labels, including capped and not converged when blocking issues remain
  • Changes left in the working tree, never committed

Common Use Cases

  • Iterating on a module until a chosen reviewer agent finds no blocking issues
  • Hardening a plugin directory with a specific review skill
  • Resuming a stopped loop after you decide how to handle an escalation
  • Running review-and-fix in a scripted session by polling the task until it ends

What Makes It Unique

It has no default reviewer and no bundled one, so the loop is only as good as the reviewer you choose. Its guards sit in the workflow: it refuses to imitate the loop inline when Workflow is unavailable, halts if the reviewer is missing, and treats a capped run as a failure to report.

Under The Hood

The skill is an instruction file that checks paths with shell commands and launches a Workflow script from the same plugin. The workflow calls your chosen reviewer and a fixer subagent, which read and edit files inside the scope you gave, and it writes a ledger and metrics to disk. It uses git to check scope after each round and may initialize a repository if none exists, which it reports as a note. It does not fetch web pages or call outside APIs by itself, though your chosen reviewer might.

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