goallisted

# Goal Use this skill when the user wants the agent to keep working toward one measurable condition instead of stopping after a single turn. ## Core Behavior `/goal` is a session-scoped autonomous workflow: 1. The user states a completion condition. 2. The agent immediately works toward that condition. 3. After each turn, evaluate whether the condition is satisfied using only evidence surfaced in the conversation. 4. If the condition is not satisfied, continue with the next concrete step. 5. Stop when the condition is met, the user clears it, or a stated bound is reached. Maintain exactly one active goal per session. If the user sets a new goal, replace the previous active goal. ## Command Handling ### Set a goal When the user writes `/goal <condition>`: - Treat `<condition>` as the directive for the next turn; do not ask for a separate prompt. - Restate the goal briefly. - Identify the proof needed to demonstrate completion. - Begin work immediately unless the condition is unsafe, impossible, or too ambiguous. Example: ```text /goal all tests in test/auth pass and the lint step is clean ``` ### Check status When the user writes `/goal` with no argument, report: - Active condition, or the most recently achieved/cleared condition. - Current status: active, achieved, cleared, blocked, or not set. - Evidence gathered so far. - Next step if active. ### Clear a goal When the user writes `/goal clear`, `/goal stop`, `/goal off`, `/goal reset`, `/goal none`, or `/goa