Run review
npx -y skills add mrge-io/skills --skill run-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Runs a local cubic AI code review using the CLI on uncommitted changes or the current branch diff. Surfaces issues grouped by priority and offers to fix them. Use when the user wants a pre-commit or pre-PR quality check, says "review my code", or asks "anything I should fix".
SKILL.md
3.3 KB, 779 tokens by cl100k_base, as published. Nobody here has run it
Run Local Code Review
Run a local cubic AI code review via CLI to catch issues before committing or opening a PR.
When To Activate
- User says "review my code", "check my changes", "run a review", or "anything I should fix"
- User is about to commit or open a PR and wants a quality check
- User asks to "scan for issues", "check for problems", or "review before merging"
- User wants to validate local changes against cubic's AI review
Inputs
- Flags (optional): Pass-through flags for the cubic CLI (e.g.,
-b main,-c HEAD~1).
Instructions
1. Check the CLI is installed
which cubic
If not found, show install options and stop:
curl -fsSL https://cubic.dev/install | bash
After installing, the user needs to authenticate with cubic auth.
2. Determine what to review
git status --porcelain
- If there are uncommitted changes ->
cubic review -j(reviews working directory) - If clean working tree ->
cubic review -b -j(reviews branch vs base) - If user provided flags ->
cubic review -j <flags>
3. Wait for review completion
After invoking cubic review, do not assume the result is final immediately.
- If output indicates queued/running status (or no final
issuesarray yet), poll until complete. - Local review window: sleep 15 seconds between retries, up to 20 retries (5 minutes total).
- GitHub/branch review window (
-bor PR-linked context): sleep 30 seconds between retries, up to 30 retries (15 minutes total). - If still incomplete after the retry window, report timeout and include the last known status.
4. Parse results
The JSON output contains an issues array. Each issue has:
priority: P0 (critical), P1 (high), P2 (medium), P3 (low)file: File pathline: Line numbertitle: Issue titledescription: Detailed explanation
5. Present issues
If no issues found, tell the user the code looks good.
If issues found, present them grouped by priority (P0 first). Highlight P0 and P1 as requiring immediate attention.
6. Offer to fix
List each issue by number so the user can pick which to fix. For each selected issue:
- Read the file and surrounding context to understand the root cause
- Validate the issue is real — if it's a false positive, explain why and skip it
- Fix it in the simplest, cleanest way possible without refactoring unrelated code
Output Format
## cubic Local Review
Found 4 issues:
### P0 — Critical (1)
1. **SQL injection in query builder** — `src/auth.ts:45`
Unsanitized user input passed directly to SQL query.
### P1 — High (1)
2. **Missing null check** — `src/api.ts:88`
`user.email` accessed without checking if user exists.
### P2 — Medium (1)
3. **Unhandled promise rejection** — `src/worker.ts:23`
Async function missing try/catch around external API call.
### P3 — Low (1)
4. **Unused import** — `src/utils.ts:1`
`lodash` imported but never used.
Which issues should I fix? (e.g., "1, 2" or "all P0 and P1")