diff --git a/core/dispatcher.py b/core/dispatcher.py
index 8c260dd..5636b9e 100644
--- a/core/dispatcher.py
+++ b/core/dispatcher.py
@@ -328,12 +328,14 @@ class AgentDispatcher:
                                 "   - Or a plan was proposed, but the human replied with feedback, corrections, or requests for changes, so we need to propose a revised plan.\n"
                                 "   You will write a detailed, step-by-step implementation plan (listing files to modify/create, specific changes to make, verification/test commands).\n"
                                 "   Your output must include a comment to post, starting with the plan and ending with a question asking if the plan is OK or if they have comments.\n"
+                                "   CRITICAL: The comment in the `comment_body` field must contain ONLY the implementation plan and the trailing confirmation question. DO NOT include your thought process, reasoning, research notes, or file analysis in the comment. Keep it concise, professional, and limited strictly to the plan itself. Any internal reasoning should be placed in the `reasoning` field of the JSON instead.\n"
                                 "   CRITICAL: The comment must contain the tags `<!-- agent:plan-proposal -->` and `<!-- agent:awaiting-reply -->` on separate lines at the very end of the comment.\n\n"
                                 "2. `ANSWER_QUESTION`: Choose this if the issue is just a question or request for information (no code changes needed), and either:\n"
                                 "   - No answer has been provided yet by the AI agent.\n"
                                 "   - Or the agent answered, but the human replied with follow-up questions or clarifications.\n"
                                 "   You will formulate a clear, helpful answer to the question.\n"
                                 "   Your output must include a comment to post, starting with the answer and ending with a question asking if this was a good enough answer.\n"
+                                "   CRITICAL: The comment in the `comment_body` field must contain ONLY the actual answer to the question and the trailing confirmation question. DO NOT include your thought process, reasoning, research notes, or file analysis in the comment. Keep it clean, concise, helpful, and limited strictly to the answer itself. Any internal reasoning should be placed in the `reasoning` field of the JSON instead.\n"
                                 "   CRITICAL: The comment must contain the tags `<!-- agent:question-response -->` and `<!-- agent:awaiting-reply -->` on separate lines at the very end of the comment.\n\n"
                                 "3. `EXECUTE_PLAN`: Choose this if:\n"
                                 "   - A plan was previously proposed (check the comment history) AND the human has clearly replied with approval/greenlight/go-ahead (e.g. 'yes', 'looks good', 'ok', 'go ahead', etc.).\n"
@@ -343,6 +345,7 @@ class AgentDispatcher:
                                 "   You will write a polite final comment to post on the issue.\n\n"
                                 "5. `NO_ACTION`: Choose this if the issue/PR is already resolved, or if we cannot proceed for another reason.\n\n"
                                 "You MUST respond ONLY with a JSON object inside a ```json markdown code block. Do not include other text.\n"
+                                "CRITICAL: The `comment_body` field in the JSON must contain ONLY the implementation plan or the answer itself, and MUST NOT contain any thought process, reasoning, or internal details. Place all thought process and internal reasoning in the `reasoning` field.\n"
                                 "Example:\n"
                                 "```json\n"
                                 "{\n"