Review Recommendations

cortex suggest --review is a standalone recommendation surface for review work. It is not the same thing as the agent-loops workflow.

The job of cortex suggest --review is:

Suggest which review-relevant skills should be loaded before you perform a review.

What It Is Not

cortex suggest --review does not run the full agent-loops review/test-audit loop for you. agent-loops is a larger implementation discipline workflow. cortex review is a composable helper you can use before review work begins.

Core Command 

cortex suggest --review --dry-run
cortex suggest --review --context debug
cortex suggest --review --context feature --context security

The command takes optional context signals and recommends review-relevant skills based on what kind of review you are about to do.

When To Use It

Use cortex suggest --review when:

  • you are about to review a diff manually
  • you want Cortex to suggest the right skills before reviewing
  • you are preparing a review-oriented handoff for another agent or model

It is especially useful when the review mode is not obvious from the diff alone and you want a nudge toward the right review discipline.

Practical Workflow 

# Ask for review-oriented skill recommendations
cortex suggest --review --context security --context api

# Build skill context for the session if needed
cortex skills context --no-write

Then load the suggested skills or use their generated slash-command aliases.

Common follow-on skills include:

  • /ctx:requesting-code-review
  • /ctx:doc-claim-validator
  • /ctx:security-testing-patterns
  • /ctx:python-testing-patterns

Relationship To Agent Loops

Keep the distinction clean:

  • cortex suggest --review recommends review-relevant skills
  • agent-loops is the full implementation workflow with review, test audit, lint, and commit gates

That means cortex suggest --review can be used on its own, or it can be one piece of a larger composable workflow.

  • Agent Loops – full implementation discipline workflow
  • Skills – inspect or rate the skills being recommended
  • Multi-LLM Consult – take review context to another model if needed