---
name: ceratops-skill-optimize
description: Optimize existing Codex or Ceratops skills by proposing targeted updates to structure, triggers, workflow, constraints, output contracts, or done criteria.
---

# Ceratops Skill Optimizer

## Goal

Propose the smallest clear skill update that improves an existing skill's structure, trigger behavior, workflow, constraints, output contract, or done criteria without changing its intended purpose unless the user explicitly requests that behavior change.

## Context

### Inputs To Capture

- Current skill text or the concrete target skill path.
- The skill's purpose, trigger conditions, expected user request, workflow, constraints, output contract, and completion criteria.
- Companion artifacts that govern the same behavior, such as `agents/openai.yaml`, bundled resources, helper references, docs, or manifests.
- Whether the requested change is structural, behavioral, trigger-only, output-only, validation-only, or wording-only.

Ask one concise question and stop if no skill text, file path, or concrete target is available.

## Constraints

### Skill-Specific Rules

- Do not apply edits or mutate files.
- Preserve the skill's existing intent unless the user explicitly asks to change behavior.
- Prefer targeted proposals over whole-skill rewrites.
- Put trigger information in YAML frontmatter descriptions rather than only in the body.
- Keep `SKILL.md` focused on execution rules that are useful after the skill has already triggered.
- Include a role only when it changes how Codex should judge tradeoffs, inspect inputs, choose tools, or decide completion.
- Inspect `agents/openai.yaml`, assets, bundled resources, and helper references for every target skill.
- When companion artifacts govern the same behavior, include the aligned companion update or state why it is intentionally retained.
- Remove or merge duplicate responsibilities and avoid overlong, brittle, vague, contradictory, or purely stylistic changes.
- Prefer deterministic scripts or helpers for repeatable procedural behavior only when text instructions are insufficient.
- Do not add auxiliary documentation, README files, examples, or unrelated resources unless they are required to prevent misinterpretation.
- A proposed update must preserve intent, improve execution clarity or trigger accuracy, have a clear owner, avoid duplicated instructions, be enforceable, define completion, and avoid unjustified recurring maintenance cost.

### Boundaries

- Use this skill for advisory skill optimization and exact proposed skill text changes.
- If the user asks to apply changes to Ceratops skills, use `$ceratops-skill-update`.
- If the user asks to create a new skill, use `$ceratops-skill-create`.
- If the requested change is one durable rule, use `$ceratops-produce-rule-update`.

### Workflow

1. Inspect the target skill text and companion metadata.
2. Identify current purpose, trigger surface, workflow, constraints, output contract, and done criteria.
3. Decide whether Goal / Context / Constraints / Done When structure would improve execution.
4. Find concrete issues: missing trigger context, unclear ownership, duplicated rules, weak completion criteria, stale labels, metadata drift, excessive output, or unverifiable instructions.
5. Propose the narrowest update that fixes the issue.
6. Include exact proposed replacement or addition and the smallest current anchor needed to locate it.

## Done When

### Completion Gate

- The response states whether an update is recommended.
- Every proposed change includes exact new text and target location.
- Behavior changes, recurring cost increases, blockers, and missing inputs are explicit.
- No files were edited.

### Output Contract

Return only:

- recommendation status
- target section or insertion point
- exact proposed replacement or addition
- minimal current anchor when needed
- behavior changes, if any
- blockers or missing inputs, if any