What it does
One paste-ready /goal prompt for Claude Code or Codex: a completion contract with checkable success criteria, so the agent works until the evidence says done or stops at a named blocker.
Who uses it
What you need
You bring:
- the task in plain words
- which tool will run it
- what proves it is done
How it works4
The steps, in order. The method itself is served whole through the connector, current on every use.
- 1
Name the target tool and the one missing fact
- 2
Draft the contract
- 3
Check every criterion is machine-checkable
- 4
Output the prompt with its character count
What it doesn't do
- Big or ambiguous tasks that need research first
- Prompts that judge on adjectives rather than commands
When your AI reaches for it
Forge a paste-ready /goal completion contract for Claude Code or Codex — fast, no research subagents. Use whenever the user wants a /goal prompt for a specific task: "write me a /goal", "create a goal prompt", "goal prompt for this", "turn this into a /goal", "make Codex build this until done", or when they describe autonomous build/fix/migrate work and want the agent to run until verified-done. Invoke with /create-goal-prompt [task] [claude-code|codex].