rinse

Orchestrates iterative review, fix, and verification loops on code diffs before PR handoff.

10|Updated May 12, 2026
One-click install
npx skills add https://github.com/sandsower/beislid --skill rinse-sandsower
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rinse
Source: https://github.com/sandsower/beislid/tree/main/skills/rinse
Command: npx skills add https://github.com/sandsower/beislid --skill rinse-sandsower

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Code review findings often pile up without a controlled process for fixing them, leading to scope creep, unverified patches, or endless patching of a fundamentally flawed design. This Skill runs a disciplined review/fix/verify loop where every fix requires user approval and every iteration is verified. ## Core Features & Use Cases - Controlled fix boundaries: Asks the user upfront which severity levels (Critical, Important, Minor) or file paths may be changed, and never expands scope silently. - Iterative review reruns: Re-invokes the side-effect-free review primitive after each fix round, comparing new findings against previous ones. - Stop and escalation conditions: Detects when repeated blocking findings in the same subsystem signal a design flaw and routes the work to redesign instead of more patching. - Use Case: Before handing off a feature branch for PR review, run the loop to triage findings, apply approved fixes, run tests, and repeat until no Critical or Important issues remain. ## Quick Start Run a rinse loop on my current branch against main, fixing only Critical and Important findings with my approval before each change.

Frequently Asked Questions about rinse

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

FAQPage Schema
How do I run an iterative code review and fix loop on a branch?▼

Invoke the rinse loop with a diff range such as a merge-base against main. It runs the review primitive, triages findings by severity with you, applies only approved fixes, runs verification, and reruns review until no blocking findings remain.

What is the difference between rinse and a single code review pass?▼

A single review pass is side-effect-free and only reports findings. Rinse is an orchestrator that edits files, runs verification, and repeats review passes within user-approved boundaries until the branch converges.

Can I limit which findings get fixed automatically?▼

Yes. Before any edits, rinse asks you to set a fix boundary: Critical only, Critical plus Important, all findings, or a custom file/path scope. It obeys that boundary for the whole loop unless you explicitly change it.

When should I stop iterating and redesign instead?▼

Stop when the same subsystem produces blocking findings for two consecutive iterations, a fix introduces new blocking findings elsewhere, or verification failures point to unclear requirements. Rinse then escalates to a blueprint/redesign step rather than more patching.

Does the review loop run tests after fixes?▼

Yes. After applying approved fixes, it runs verification appropriate to the changed files, preferring configured project gates or inferred checks like unit tests, typecheck, build, and lint. Failed verification is investigated before further code changes.