skills.rest
Norfolk-Group
Norfolk-GroupOFFICIAL
Explore all skills by this author →
GitHub

ce-code-review

Reviews code changes with parallel persona agents and merges findings into a severity-ranked report.

Updated Mar 17, 2026
Search Tags:#static analysis#code review#pull request#github#multi-agent#autofix#severity
One-click install
npx skills add https://github.com/Norfolk-Group/marcela-norfolk-ai --skill ce-code-review-norfolk-group
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ce-code-review
Source: https://github.com/Norfolk-Group/marcela-norfolk-ai/tree/main/skills/compound-engineering/skills/ce-code-review
Command: npx skills add https://github.com/Norfolk-Group/marcela-norfolk-ai --skill ce-code-review-norfolk-group

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Manual code review is slow and inconsistent, and single-pass AI review misses entire classes of defects. This Skill runs a structured multi-agent review that dispatches specialized reviewer personas in parallel, gates findings by confidence, and merges them into one deduplicated report before you open a PR. ## Core Features & Use Cases - Tiered Persona Reviewers: Spawns always-on reviewers (correctness, testing, maintainability, project standards) plus conditional reviewers for security, performance, API contracts, migrations, and stack-specific concerns like Rails, Python, TypeScript, and Swift. - Confidence-Gated Findings: Every finding carries a P0-P3 severity, an anchored confidence score (0/25/50/75/100), and an autofix routing class, so low-signal noise is suppressed before it reaches you. - Four Run Modes: Interactive, autofix (applies safe fixes automatically), report-only (strictly read-only), and headless (structured output for programmatic callers). - Use Case: Before creating a PR for a Rails auth feature, run the review to get security, reliability, and Rails-convention findings merged into a single table with a verdict, fix order, and residual work routed to tickets. ## Quick Start Ask the agent to review the current branch with ce-code-review before creating a pull request.

Frequently Asked Questions about ce-code-review

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

FAQPage Schema
How do I run a multi-agent code review before creating a PR?▼

Invoke the review with no arguments to review the current branch, or pass a PR number, GitHub URL, or branch name. The skill computes the diff against the merge-base, spawns reviewer personas in parallel, and returns a merged severity-ranked report.

What reviewer personas does this code review use?▼

Every run spawns correctness, testing, maintainability, and project-standards reviewers plus agent-native and learnings agents. Conditional reviewers for security, performance, API contracts, migrations, reliability, and stack-specific stacks like Rails, Python, TypeScript, and Swift are added based on the diff.

Can the code review apply fixes automatically?▼

Yes, autofix mode applies only findings classified as safe_auto, such as missing nil guards or dead code removal, in bounded re-review rounds. It never commits, pushes, or creates PRs, and writes a run artifact summarizing residual work.

Does the review work in CI or headless pipelines?▼

Headless mode supports programmatic callers by applying safe fixes in a single pass and returning all other findings as structured text with a run artifact path. It requires a determinable diff scope and ends with a Review complete signal.

Why are some review findings suppressed from the report?▼

Findings below confidence anchor 75 are suppressed unless they are P0 severity, since anchors 0 and 25 indicate unverifiable or false-positive issues. The Coverage section of the report lists how many findings were suppressed and why.

When should I use report-only mode instead of autofix?▼

Use report-only when you need a strictly read-only review with no edits, artifacts, or commits, such as when running concurrently with browser testing on the same checkout. It is the only mode safe for parallel read-only verification.

Related Skills

Explore More Software Engineering Skills→
norrbackanorrbacka

requesting-code-review

Review completed code changes against requirements before merge.

Community
Intermediate
SeanHoggSeanHogg
3

prepare-pr

Rebase a GitHub Pull Request onto main, fix review findings, run gates, and push.

Community
Advanced
StarryCodStarryCod
11

requesting-code-review

Run static security scans, tests, linting, and independent review on staged git changes.

Community
Advanced
thilthpc01208thilthpc01208

quick-design

Generate lightweight design briefs and embed them into story files.

Community
Basic