How does this differ from just asking a standard AI model to review my code?

Unlike general LLMs that offer conversational feedback, this skill uses a structured severity framework (Critical to Style) and a pre-action gate to filter out fluff, ensuring every comment is anchored to a specific line and a concrete fix.

Which programming languages and platforms are supported?

This skill is language-agnostic and works with any programming language, though it is specifically optimized for Git workflows and GitHub PR URLs to provide the most contextually aware feedback.

How much setup is required to align the skill with my project's coding standards?

Once installed, simply use the `maji-review` command followed by a file path, directory, or PR URL; the skill automatically analyzes your project's configuration files to understand your specific coding conventions.

Can I customize the level of detail or the severity threshold for the reviews?

By design, the skill filters out minor 'nitpicks' and focuses on high-impact findings, though you can adjust the focus by targeting specific files or diffs to ensure the output remains relevant to your current task.

Does the skill require special permissions or access to my entire repository?

The skill requires access to your local environment or a public/authenticated PR URL to analyze files, but it does not store your code or use it for training purposes beyond the immediate review session.

Redmark: code review by severity

Example session with this skill installed

maji-review src/db.ts (contains raw string interpolation in a query and a missing try/catch on a connection string)

Read your context and instructions
Compiled the redmark

2 findings: 🔴 1 critical, 🟡 1 important.

🔴 Critical (1)
──────────────────────
src/db.ts:12: SQL injection via dynamic string
Why: Database query uses template literals for user input without sanitization.
Fix: db.execute('SELECT * FROM users WHERE id = ?', [userId])

🟡 Important (1)
──────────────────────
src/db.ts:5: Unhandled connection error
Why: The connection string is read from env without a try/catch or null check, causing immediate crash if missing.

Fix

try {
  await client.connect();
} catch (e) {
  logger.error('DB_CONN_FAIL', e);
  process.exit(1);
}

Verdict: Block merge. Fix critical SQL injection and add connection error handling.

Next steps

Implement parameterized queries in all database utility files.
Add environment variable validation at the application entry point.

Connects securely to your tools. The creator never sees your data.

About this skill

Code reviews often suffer from two extremes: vague, flowery praise that hides bugs, or overwhelming walls of text that are impossible to triage. This skill enforces a disciplined, evidence-first review process that categorizes every finding by severity and provides immediate code fixes.

What it does

Severity-based grouping categorizes issues into Critical, Important, Style, and Suggestion buckets for faster triaging.
Evidence-first reporting maps every finding to a specific file:line reference with a one-sentence justification.
Actionable fixes provides concrete code snippets or pattern references for every identified issue instead of abstract advice.
Noise reduction filters out "Great work!" preambles and repetitive comments, focusing strictly on meaningful findings.
Scope enforcement prevents sweeping refactor suggestions when reviewing single files or narrow diffs.

How it works

Context gathering analyzes your README, package manifests, and config files to understand your stack and conventions.
Targeting invokes the review using Redmark <target>, where the target is a file path, a directory, a git diff, or a PR URL.
Automated triage runs findings through a pre-action gate to ensure suggestions are within scope and high-impact.
Structured output delivers a summary line, grouped findings with fixes, and a final merge verdict.

Frameworks & tools

Works with any programming language or framework. Optimized for git workflows, PR URLs (GitHub), and local directory structures.

Why this beats prompting it yourself

Standard LLM reviews are prone to "hallucinating" politeness or suggesting massive refactors that break dependencies. This skill uses a strict severity framework and a pre-action gate to ensure every comment is anchored to a line of code and a specific risk level.

Use cases

Pre-flight checks to catch SQL injection, race conditions, or plain-text secrets before a git push.
PR triaging to generate a structured audit of a team member's pull request URL.
Legacy cleanup to identify tight coupling and performance bottlenecks in specific existing files.

Known limitations

Does not support commenting on every line by design. Requires local context or provided URLs to evaluate dependencies accurately.

How to use

Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

Reviews

No reviews yet

Be one of the first to try it. Every listed skill passes our trust checks below.

Security scanned

Passed our 8-point scan before listing

1 install

Downloaded by developers to date

Free forever

No account required to browse

Redmark: code review by severity