smart-review-fix

Orchestrates a closed-loop pipeline that reviews code, creates fix stories, executes repairs, and verifies results.

10|3|Updated Mar 7, 2026
One-click install
npx skills add https://github.com/Cynthia1070711/PHYCOOL_Tools --skill smart-review-fix-cynthia1070711
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: smart-review-fix
Source: https://github.com/Cynthia1070711/PHYCOOL_Tools/tree/main/config-templates/claude/skills/smart-review-fix
Command: npx skills add https://github.com/Cynthia1070711/PHYCOOL_Tools --skill smart-review-fix-cynthia1070711

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires better-sqlite3, and includes scripts (resource) and references (resource) components.

What problem does it solve? Code review findings often get lost between discovery and verification, leaving bugs untracked or falsely marked as fixed. This Skill closes that gap by orchestrating a five-phase pipeline (Discover, Analyze, Create, Fix, Verify) that carries every finding from detection through code-level verification, with a database-backed state machine tracking each bug's status. ## Core Features & Use Cases - Full-Spectrum Review Orchestration: Launches batch code, security, and E2E review sub-windows via PowerShell scripts, writing findings into a SQLite review_findings table. - Automated Story Creation & Scheduling: Groups findings into fix Stories, detects file conflicts and dependencies, and dispatches parallel fix pipelines with dynamic slot backfilling. - E2E Loop Mode: Drives browser-based exploratory testing through Chrome DevTools MCP sub-windows, then loops fix-and-retest cycles (up to 3 rounds per bug) until PASS or deferred. - Verified Backfill: Requires reading actual code at file:line before marking findings fixed, preventing false-positive closures. - Use Case: Run a full review of the editor and payment modules, automatically group the 30 discovered bugs into prioritized Stories, dispatch parallel fix pipelines, then verify each fix against real code and produce a closed-loop report with fix-rate statistics. ## Quick Start Ask the agent to run a full smart review-fix cycle on the editor-core and payment modules and report the verified fix rate.

Frequently Asked Questions about smart-review-fix

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

FAQPage Schema
How do I run a full automated code review and fix pipeline?

Invoke the full mode with target modules, for example /smart-review-fix full editor-core,payment. The pipeline runs batch code, security, and E2E reviews, groups findings into Stories, dispatches fix pipelines, and verifies each fix against actual code.

How does the E2E loop mode test and fix bugs automatically?

E2E loop mode launches a Sonnet sub-window that drives Chrome DevTools MCP through functional paths, records structured bug reports, then dispatches fix sub-windows and re-tests each fix. Bugs failing three repair rounds are marked deferred.

Can multiple fix pipelines run in parallel?

Yes, general fixes run up to 2-3 concurrent sub-windows with staggered 10-second launches. Stories sharing files are serialized, and E2E or Chrome MCP tasks are limited to one at a time because Chrome allows a single connection.

Why must fixes be verified by reading code instead of story status?

Story status alone produces roughly a 33 percent false-positive rate, so the pipeline requires reading the actual file and line referenced by each finding. Only after code evidence confirms the fix is the finding marked fixed in the database.

What happens to bugs that conflict with intentional design decisions?

During analysis the pipeline checks a memory database for deliberate business rules, such as Free plan gating or refund policy. Conflicting findings are marked wont_fix with an [INTENTIONAL] note explaining the reason and reference.