titans_skill

Overview

This skill runs a three-lens code review using parallel subagents: Epimetheus (hindsight), Metis (craft), and Prometheus (foresight). Triggered by /titans, /review, or phrases like "review this code", it inspects scoped work after substantial changes and produces a synthesized action-oriented summary. Use it as the final quality gate before /close.

How this skill works

You scope the review (files, directories, or session changes) and dispatch three parallel reviewers using deep reasoning models. Each reviewer returns structured findings (findings, assumptions, could-not-assess, questions). The skill synthesizes overlaps, conflicts, and a prioritized critical path into a single actionable report.

When to use it

• After completing substantial work, before calling /close
• As the final quality gate prior to shipping a feature or release
• Periodic hygiene to uncover hidden debt or rot
• When you need multiple perspectives (bugs, craft, and future fit)
• When the change is long-lived or consumed by other systems

Best practices

• Scope narrowly and explicitly — avoid sending the entire repo accidentally
• Provide context (CLAUDE.md, README, intended consumers) to reduce 'could not assess' gaps
• Run only for substantial, long-lived changes — skip tiny fixes or throwaway prototypes
• Dispatch all three reviewers in parallel and always synthesize their outputs
• If a reviewer fails, rerun with tighter scope and note the missing lens in the synthesis

Example use cases

• Final review of a completed feature branch before merging
• Evaluate a refactor for regressions, clarity, and future extensibility
• Assess documentation, architecture notes, or process changes using the same three-lens pattern
• Periodic audit of a subsystem to find hidden fragility and roadmap gaps
• Pre-release checklist to surface critical fixes and unresolved questions