Write Good Goal

Invoke as /afk:write-good-goal. Use it when you need to draft, tighten, validate, or repair a /goal command, completion condition, or success criteria for long-running agent work.

write-good-goal turns vague intent into a single agent-evaluable completion condition — shaped, then checked against six quality criteria.

What it does

Write Good Goal turns vague intent into a single, agent-evaluable completion condition. It identifies the intended outcome, converts it to a measurable condition an agent can verify from its own transcript, and presents the result without starting the work. For frontend outcomes it includes real browser verification via agent-browser; for backend, CLI, or docs work it uses the narrowest contract, test, or artifact proof that fits.

• Uses the shape: Achieve [specific outcome]. Prove it by [verification]. Preserve [constraints]. Stop if [limit].
• Checks the result against six quality criteria: specific, measurable, agent-observable, evidence-shaped, bounded, non-ambiguous.
• Presents the recommended /goal and explains the assumptions; offers a tighter variant when a safer bounded alternative exists.

View the full skill on GitHub