skills.rest
zts212653
zts212653COMMUNITY
Explore all skills by this author →
GitHub

fresh-context-review

Generates signed finding lists from PR diffs before formal code review.

2.9k|733|Updated Mar 12, 2026
Search Tags:#Git#quality gate#pre-review#Code review#diff scan#Pull request#finding list
One-click install
npx skills add https://github.com/zts212653/clowder-ai --skill fresh-context-review
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fresh-context-review
Source: https://github.com/zts212653/clowder-ai/tree/main/cat-cafe-skills/fresh-context-review
Command: npx skills add https://github.com/zts212653/clowder-ai --skill fresh-context-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Formal reviewers face high cognitive load when reviewing large pull requests cold. This Skill runs an author-triggered fresh-context scan of the PR diff before formal review, producing a finding list that lets reviewers focus on deeper issues instead of surface-level problems.

Core Features & Use Cases

  • Pre-Review Diff Scanning: A fresh-context agent (a teammate not involved in development, or a new session) reads the PR diff and spec, then reports findings across correctness, security, performance, naming, spec-mismatch, and missing-test categories.
  • Trigger Decision Table: Built-in guidance on when to run the scan (multi-file changes, shared/ cross-package edits, state machine changes) and when to skip it (docs-only, trivial diffs, hotfixes).
  • Structured Finding List Output: Produces a signed markdown table with file, line, category, severity (P1/P2/P3), and totals, explicitly without any APPROVE/BLOCK verdict.
  • Use Case: After passing the quality gate on a branch touching five files and 200 lines, the author asks a cross-model teammate for a fresh-context scan, fixes two P2 findings, and attaches the finding summary to the formal review request.

Quick Start

Ask a teammate who did not participate in development to run a fresh-context pre-review scan on your branch diff and return only a finding list in the specified format.

Frequently Asked Questions about fresh-context-review

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

FAQPage Schema
How do I run a pre-review scan on a pull request diff?▼

After your quality gate passes, ask a teammate who did not participate in development to review the output of git diff origin/main...HEAD along with your spec. They return only a finding list with file, line, category, and severity, not a verdict.

When should I use a fresh-context review before formal review?▼

Use it for PRs with three or more files and 50 or more diff lines, shared or cross-package changes, and state machine or lifecycle changes. Skip it for docs-only changes, diffs under 10 lines, SKILL.md-only edits, and urgent hotfixes.

Does a fresh-context scan replace formal code review approval?▼

No. The fresh-context scan is strictly a finding generator with zero approval authority. It cannot produce APPROVE or BLOCK verdicts, does not affect merge gates, and the formal request-review process with a named reviewer is still required.

Can I do a fresh-context review of my own code?▼

Yes, as a fallback you can open a brand new session and scan your own diff, since the new session lacks development context. However, cross-model or cross-teammate scans yield more findings because different model families have different blind spots.

What severity levels do fresh-context findings use?▼

Findings use three severity levels: P1 for runtime errors, data loss, or security holes; P2 for incomplete logic, missing tests, or spec mismatches; P3 for naming, style, and documentation issues. The author decides which findings to fix or dismiss.

Related Skills

Explore More Software Engineering Skills→
multiplex-aimultiplex-ai
2

verification-loop

Run build, type-check, lint, test, security, and diff checks into a readiness report.

Official
Advanced
LightForgeLabsStudioLightForgeLabsStudio

pr-review

Review pull requests against issue specifications and quality standards.

Official
Intermediate
recallnetrecallnet
1

verify

Verify in-repo payloads via cross-model reviewers and update prosecution documents.

Official
Advanced
ajayjoshi-ciphersonicajayjoshi-ciphersonic

requesting-code-review

Automate code review with static security scanning and baseline-aware quality checks.

Community
Advanced