Exact request sent to the run panel.
{
"source": "retrieve",
"template": {
"id": "x_commenter",
"slug": "x-commenter",
"name": "X/Twitter Commenter",
"categoryId": "SOCIAL_ENGAGE"
},
"inputs": {
"tweetUrl": "",
"replyText": "",
"replyStyle": "smart",
"replyAngle": "",
"autoPost": true
},
"agentHint": "You are replying to a tweet on X/Twitter. Adapt to the actual UI you encounter.\n\n**TASK CONFIGURATION:**\n- Tweet URL: \n- User's Reply: \n- Style: smart\n- Angle: \n- Auto-post: true\n\n---\n\n## STEP 1: NAVIGATE TO TWEET\n\n1. Go to \n2. Wait for tweet to fully load\n3. If not logged in, STOP and report authentication required\n4. If tweet not found/deleted, report error\n\n---\n\n## STEP 2: READ & UNDERSTAND THE TWEET\n\n**Analyze the tweet:**\n- Full tweet content\n- Author and their context\n- Any media, links, or quoted tweets\n- Thread context (is this part of a thread?)\n- Existing replies (note the conversation tone)\n- Engagement level (likes, retweets, replies)\n\n---\n\n## STEP 3: PREPARE REPLY\n\n### IF user provided replyText:\nUse this reply exactly:\n```\n\n```\n\n**VALIDATION:**\n- X replies have a 280 character limit\n- If reply exceeds 280 chars: STOP and report \"Reply exceeds 280 character limit ( chars). Please shorten.\"\n- Count characters accurately (URLs count as 23 chars)\n\n### IF replyText is empty - GENERATE REPLY:\n\n**Generate a smart reply that:**\n\n1. **Responds to their specific point**\n - Reference what they actually said\n - Don't just agree generically\n\n2. **Follows the angle:**\n \"\"\n\n3. **Matches the style:**\n - smart: Insightful observation, add perspective\n - witty: Clever, funny (but not try-hard)\n - question: Genuine question that invites response\n - data: Add relevant fact, stat, or source\n - contrarian: Respectful pushback with reasoning\n - supportive: Amplify with specific agreement\n\n4. **X reply format:**\n - Punchy and concise\n - Under 280 characters (ideally under 200)\n - NO hashtags (looks desperate in replies)\n - Minimal emojis unless the vibe calls for it\n\n---\n\n## STEP 4: POST OR PREVIEW\n\n1. Find the reply input area\n2. Click to focus (may need to click \"Reply\" first)\n3. Type or paste the reply\n\n### IF true === true:\n4. Find and click the reply/post button\n5. Wait for reply to appear in thread\n6. Capture reply URL if possible\n\n### IF true === false:\n4. Leave reply in input field\n5. Report ready for review\n\n---\n\n## STEP 5: REPORT RESULTS\n\n```json\n{\n \"success\": true|false,\n \"originalTweet\": \"<summary of the tweet>\",\n \"tweetAuthor\": \"\",\n \"replyText\": \"<the reply used>\",\n \"replySource\": \"provided|generated\",\n \"characterCount\": <number>,\n \"withinLimit\": true|false,\n \"status\": \"posted|ready_for_review|failed|over_limit\",\n \"replyUrl\": \"<if posted>\",\n \"errorMessage\": \"<if failed>\"\n}\n```\n\n---\n\n## CHARACTER LIMIT HANDLING\n\n- STRICT 280 character limit on X\n- Always count characters before posting\n- If over limit: DO NOT truncate without user consent\n- Report exact count and ask user to revise\n\n## ADAPTIVE UI HANDLING\n\n- Reply button may be icon or text\n- Reply compose area may be inline or modal\n- Thread context affects how reply appears",
"userInput": "You are replying to a tweet on X/Twitter. Adapt to the actual UI you encounter.\n\n**TASK CONFIGURATION:**\n- Tweet URL: \n- User's Reply: \n- Style: smart\n- Angle: \n- Auto-post: true\n\n---\n\n## STEP 1: NAVIGATE TO TWEET\n\n1. Go to \n2. Wait for tweet to fully load\n3. If not logged in, STOP and report authentication required\n4. If tweet not found/deleted, report error\n\n---\n\n## STEP 2: READ & UNDERSTAND THE TWEET\n\n**Analyze the tweet:**\n- Full tweet content\n- Author and their context\n- Any media, links, or quoted tweets\n- Thread context (is this part of a thread?)\n- Existing replies (note the conversation tone)\n- Engagement level (likes, retweets, replies)\n\n---\n\n## STEP 3: PREPARE REPLY\n\n### IF user provided replyText:\nUse this reply exactly:\n```\n\n```\n\n**VALIDATION:**\n- X replies have a 280 character limit\n- If reply exceeds 280 chars: STOP and report \"Reply exceeds 280 character limit ( chars). Please shorten.\"\n- Count characters accurately (URLs count as 23 chars)\n\n### IF replyText is empty - GENERATE REPLY:\n\n**Generate a smart reply that:**\n\n1. **Responds to their specific point**\n - Reference what they actually said\n - Don't just agree generically\n\n2. **Follows the angle:**\n \"\"\n\n3. **Matches the style:**\n - smart: Insightful observation, add perspective\n - witty: Clever, funny (but not try-hard)\n - question: Genuine question that invites response\n - data: Add relevant fact, stat, or source\n - contrarian: Respectful pushback with reasoning\n - supportive: Amplify with specific agreement\n\n4. **X reply format:**\n - Punchy and concise\n - Under 280 characters (ideally under 200)\n - NO hashtags (looks desperate in replies)\n - Minimal emojis unless the vibe calls for it\n\n---\n\n## STEP 4: POST OR PREVIEW\n\n1. Find the reply input area\n2. Click to focus (may need to click \"Reply\" first)\n3. Type or paste the reply\n\n### IF true === true:\n4. Find and click the reply/post button\n5. Wait for reply to appear in thread\n6. Capture reply URL if possible\n\n### IF true === false:\n4. Leave reply in input field\n5. Report ready for review\n\n---\n\n## STEP 5: REPORT RESULTS\n\n```json\n{\n \"success\": true|false,\n \"originalTweet\": \"<summary of the tweet>\",\n \"tweetAuthor\": \"\",\n \"replyText\": \"<the reply used>\",\n \"replySource\": \"provided|generated\",\n \"characterCount\": <number>,\n \"withinLimit\": true|false,\n \"status\": \"posted|ready_for_review|failed|over_limit\",\n \"replyUrl\": \"<if posted>\",\n \"errorMessage\": \"<if failed>\"\n}\n```\n\n---\n\n## CHARACTER LIMIT HANDLING\n\n- STRICT 280 character limit on X\n- Always count characters before posting\n- If over limit: DO NOT truncate without user consent\n- Report exact count and ask user to revise\n\n## ADAPTIVE UI HANDLING\n\n- Reply button may be icon or text\n- Reply compose area may be inline or modal\n- Thread context affects how reply appears",
"urls": [],
"meta": {
"createdAt": "",
"userAgent": ""
},
"execution": {
"mode": "device",
"deviceIds": []
}
}
Add to Chrome