Toolkit
Commands & Workflows · Playbook
On this page7 sections
Playbook
1. Define Intent and Inputs
Use a verb-object name and a narrow input schema. Reject ambiguous destinations before any side effect.
2. State Authority Boundaries
List which local changes are authorized by invocation and which external or irreversible actions still require confirmation.
3. Put Complexity in a Skill
Use a thin command entrypoint for manual timing. Keep multi-step logic, references, and scripts in a reusable skill when the product supports it.
4. Make Verification First-Class
The workflow must name the decisive checks and return their actual results. “Done” without evidence is a failed contract.
5. Add Product Adapters
Map the same workflow to each supported product's current command and skill syntax. Document surfaces where no equivalent adapter has been verified.
6. Test Safety
Exercise missing inputs, invalid paths, denied tools, interrupted execution, resumed sessions, duplicate invocation, and an attempted external mutation without approval.
7. Version the Workflow
When output or authority changes materially, update the workflow version and its tests so evidence can be interpreted later.
Source register
Commands & Workflows sources
Primary vendor documentation reviewed on the date shown. Links may change after publication.
- Extend Claude with skills (opens in a new tab)
Anthropic · official docs · accessed 2026-08-01
- Claude Code commands (opens in a new tab)
Anthropic · official docs · accessed 2026-08-01
- Build skills (opens in a new tab)
OpenAI · official docs · accessed 2026-08-01
- Codex slash commands (opens in a new tab)
OpenAI · official docs · accessed 2026-08-01
- GitHub Copilot CLI command reference (opens in a new tab)
GitHub · official docs · accessed 2026-08-01