| name | workflow-to-skill |
| description | Use when turning a repeated agent workflow, prompt sequence, operating procedure, or domain process into a concise Codex or Claude skill. |
Workflow To Skill
Use this skill when a repeated workflow should become an installable skill instead of living as a one-off prompt.
Workflow
- Identify the trigger: when should the skill activate?
- Identify the user outcome: what should the agent produce?
- Remove generic advice the model already knows.
- Keep the main
SKILL.md concise.
- Move long examples or reference material into
references/ only if needed.
- Add scripts only when deterministic execution matters.
- Include output formats that make success easy to recognize.
- Add guardrails for risky decisions.
Skill Skeleton
---
name: skill-name
description: Use when ...
---
# Skill Name
## When To Use
## Workflow
## Output Format
## Guardrails
Quality Checklist
- The description includes the real trigger phrases and task types.
- The body is short enough to load often.
- The skill gives degrees of freedom instead of forcing one brittle path.
- The output format matches the user's next action.
- The skill avoids unrelated documentation files inside the skill folder.
- The examples are realistic, not toy prompts.
Anti-Patterns
- A skill that is just a long prompt.
- A skill that explains concepts the model already knows.
- A skill that silently changes secrets, credentials, git history, or production systems.
- A skill with ten outputs when the user needs one decision.
Context
This is a meta-skill. Use it at any point in the pipeline when a repeated workflow should become an installable skill instead of a one-off prompt.