review-to-disk

Save multi-agent code review reports to date-stamped directories.

2|Updated Feb 23, 2026
One-click install
npx skills add https://github.com/roach88/claude-config --skill review-to-disk
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: review-to-disk
Source: https://github.com/roach88/claude-config/tree/main/skills/review-to-disk
Command: npx skills add https://github.com/roach88/claude-config --skill review-to-disk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Standard multi-agent code reviews generate large reports that overflow the orchestrator's context window before all results can be processed, leading to lost review data and incomplete audits of codebases.

Core Features & Use Cases

  • Parallel Specialized Agents: Runs multiple background review agents focused on architecture, security, performance, code quality, and other domains, with agent rosters tailored to TypeScript, Rails, or Python projects.
  • Persistent Report Storage: All agent reports are automatically saved to date-stamped docs/reviews/YYYY-MM-DD/ directories, ensuring no review data is lost even if context limits are hit.
  • Unified Synthesis: A dedicated sub-agent compiles all individual reports into a single deduplicated, severity-sorted synthesis report for easy consumption.
  • Use Case: Run a full audit of your src/ directory before a major release to catch security vulnerabilities, performance bottlenecks, and code quality issues across your entire codebase.

Quick Start

Use the review-to-disk skill to run a full exhaustive code review of your project's source code directory and save all reports to persistent disk storage.

Frequently Asked Questions about review-to-disk

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

FAQPage Schema
How do I prevent lost code review reports when auditing large codebases?

Prevent lost code review reports by saving multi-agent review outputs directly to date-stamped disk directories. This ensures no review data is lost even if the orchestrator's context window overflows during large codebase audits.

What is the best way to run an exhaustive multi-agent code review without context window overflow?

Avoid context window overflow by running parallel specialized background agents for architecture, security, and performance, then saving all individual reports to disk before a dedicated sub-agent compiles a unified synthesis.

Can I run multi-agent source code audits for TypeScript, Rails, and Python projects?

Yes, you can run multi-agent source code audits for TypeScript, Rails, and Python projects. The review process uses agent rosters specifically tailored to each of these tech stacks to evaluate code quality and security.

How do I compile multiple background agent reports into a single code audit summary?

Compile multiple background agent reports into a single code audit summary using a dedicated sub-agent that processes individually saved reports from disk to create a deduplicated, severity-sorted synthesis report for easy consumption.

Why does my multi-agent code review drop reports before finishing a large source directory audit?

Multi-agent code reviews drop reports because large codebase audits generate massive outputs that exceed the orchestrator's context window limit. Persisting these reports to disk storage prevents this data loss.

Does the synthesized code review report get saved to disk with the individual agent findings?

Yes, the synthesized code review report is saved to disk alongside individual agent findings. All reports are automatically stored in date-stamped docs/reviews/YYYY-MM-DD/ directories to ensure persistent output.