Merge pull request #7175 from SparkofSpike/codex/goal-yield-on-question

fix(goal): name the ask-the-user exit in continuation prompts

crates/tui/src/prompts/text.rs

@@ -187,10 +187,13 @@ state — files, command output, tests, runtime behavior, issue or PR state, or
 other authoritative evidence — then call `update_goal` with
 `status: "complete"` and concise evidence. If something genuinely prevents
 progress, call `update_goal` with `status: "blocked"` and explain it. If you
-finished a stage and the next step is the user's call, call `update_goal` with
-`status: "yield"`, say what you need from them, and end your answer there;
-their reply resumes the goal. If `update_goal` is not in your tool list, load
-it with `tool_search` first.
+finished a stage and the next step is the user's call, or you need the user's
+answer before the next pass can do anything useful, call `update_goal` with
+`status: "yield"`, ask what you need in your answer, and end there; their reply
+resumes the goal. Asking in prose without calling `update_goal` does not pause
+this goal: the next pass starts with your question still unanswered, so do not
+invent an answer to keep the loop moving. If `update_goal` is not in your tool list,
+load it with `tool_search` first.
 "#;
 /// Memory hygiene guidance — appended to the system prompt only when the
 /// session has a non-empty user-memory block. Steers the model toward

crates/tui/src/tools/goal.rs

@@ -820,7 +820,7 @@ pub fn thread_goal_status_projection(
 pub fn render_continuation_prompt(snapshot: &GoalSnapshot, continuation_index: u32) -> String {
     let goal_json = serde_json::to_string_pretty(snapshot).unwrap_or_else(|_| "{}".to_string());
     format!(
-        "{}\n\n## Active Goal State\n\n```json\n{}\n```\n\nContinuation pass #{}.\nIf a critical verifier finds remaining work, call `update_goal` with `status: \"not_achieved\"` and its concrete `verification.gaps`; {} equivalent gap sets in a row pause this goal (`no progress`) for inspection instead of spending indefinitely, so report what actually still fails rather than restating the previous pass. If the goal is complete, first run or cite a concrete verifier/check when one applies, then call `update_goal` with `status: \"complete\"`, concrete evidence, and `verification: {{\"status\":\"passed\",\"check\":\"...\",\"summary\":\"...\"}}`. For non-verifiable work (docs, research, writing), use `verification: {{\"status\":\"not_applicable\",\"check\":\"...\",\"summary\":\"...\"}}` with a clear rationale instead of fabricating a verifier receipt. If it is blocked, call `update_goal` with `status: \"blocked\"` and the blocker. Otherwise continue making progress toward the objective.",
+        "{}\n\n## Active Goal State\n\n```json\n{}\n```\n\nContinuation pass #{}.\nIf a critical verifier finds remaining work, call `update_goal` with `status: \"not_achieved\"` and its concrete `verification.gaps`; {} equivalent gap sets in a row pause this goal (`no progress`) for inspection instead of spending indefinitely, so report what actually still fails rather than restating the previous pass. If the goal is complete, first run or cite a concrete verifier/check when one applies, then call `update_goal` with `status: \"complete\"`, concrete evidence, and `verification: {{\"status\":\"passed\",\"check\":\"...\",\"summary\":\"...\"}}`. For non-verifiable work (docs, research, writing), use `verification: {{\"status\":\"not_applicable\",\"check\":\"...\",\"summary\":\"...\"}}` with a clear rationale instead of fabricating a verifier receipt. If it is blocked, call `update_goal` with `status: \"blocked\"` and the blocker. If the next pass cannot make useful progress without an answer or decision only the user can give, call `update_goal` with `status: \"yield\"`, ask for it, and stop; do not invent the answer and continue. Otherwise continue making progress toward the objective.",
         crate::prompts::GOAL_CONTINUATION_PROMPT.trim(),
         goal_json,
         continuation_index,