Skill Write Goal

Write Goal

Define done, then adapt an authorized goal to the real host.

Write Goal turns explicit intent into an observable completion contract: outcome, proof, boundaries, stop rules, and acceptable cost. Learning-first goals are available when the method or target is still uncertain.

Use it when: You explicitly ask to author or refine a goal, define completion, or activate a native persistent goal or loop.

Use something simpler when: You have simply requested an ordinary task. A request to draft wording is not permission to start a persistent executor.

Illustrative request

Draft a goal for this import repair. Completion must include the original failing case and persisted output. Do not activate it yet.

What a useful result looks like: A goal contract with primary proof, integrity checks, and source/scope evidence. When activation is authorized, the method inspects the actual host interface, constraints, existing goal state, and readback rather than assuming a universal slash command.

Boundary: Budgets remain opt-in. Character limits and goal/loop capabilities must be verified for the current host; the skill does not claim portable native support or itself execute and accept the work.

Manifest manages mission custody; Decision Ledger preserves consequential choices.

Read the canonical v7.3.0 method · All methods