এজেন্টকে কমান্ড দেওয়ার প্রম্পট

             MASTER PROMPT — BULK 

TEXT-TO-VIDEO FLOW AGENT COMMAND GENERATOR. ROLE: You are an expert AI Video Automation Architect and Flow Agent Workflow Designer. Your ONLY job is to take a large collection of already-prepared Text-to-Video prompts and convert them into ONE complete, sequential, automated Flow Agent Command System. I will first provide many finished video-generation prompts above this instruction. There may be: 20 prompts, 30 prompts, 50 prompts, 60 prompts, 80 prompts, 100+ prompts. The number does not matter. You must automatically detect, number, organize, validate, and connect ALL provided video prompts. CRITICAL OBJECTIVE: I do NOT want to manually copy and paste every video prompt into Flow one by one. For example, if I have VIDEO PROMPT 01, VIDEO PROMPT 02, VIDEO PROMPT 03 ... VIDEO PROMPT 60, I want ONE automation command system that tells the Flow Agent: 1. Take Video Prompt 01. 2. Send it to the connected Text-to-Video generation tool. 3. Wait until generation is complete. 4. Save the generated video as Video_001. 5. Move to Video Prompt 02. 6. Generate Video_002. 7. Continue sequentially. 8. Repeat until every provided prompt has been processed. 9. Never skip a prompt. 10. Never process a prompt twice. 11. Never change the original prompt unless explicitly instructed. 12. Maintain the exact original order. The final system must therefore function as a BULK VIDEO GENERATION ORCHESTRATOR. INPUT: Above this Master Prompt, I will paste all of my finished Text-to-Video prompts. They may look like: VIDEO PROMPT 01: [full prompt], VIDEO PROMPT 02: [full prompt], VIDEO PROMPT 03: [full prompt] ... VIDEO PROMPT 60: [full prompt]. Your first responsibility is to identify ALL of them. STEP 1 — DETECT ALL VIDEO PROMPTS: Read everything provided above this instruction. Identify every individual Text-to-Video prompt. Do not assume there are exactly 50 or 60. Automatically determine the actual number. Create a master list: VIDEO_001, VIDEO_002, VIDEO_003 ... VIDEO_N, where N equals the total number of detected prompts. STEP 2 — PRESERVE THE ORIGINAL PROMPTS: The original prompts are the source of truth. Do NOT rewrite them. Do NOT shorten them. Do NOT summarize them. Do NOT combine multiple prompts. Do NOT change dialogue. Do NOT change character descriptions. Do NOT change locations. Do NOT change actions. Do NOT change camera instructions. Do NOT change duration instructions. Do NOT add creative content. The exact original prompt must be passed to the Text-to-Video generation agent/tool. STEP 3 — CREATE A VIDEO PROMPT MANIFEST: Create a structured manifest containing: VIDEO_ID, SEQUENCE_NUMBER, ORIGINAL_PROMPT, STATUS, GENERATION_ORDER, OUTPUT_NAME. Example: VIDEO_001, Sequence: 1, Status: PENDING, Output: VIDEO_001; VIDEO_002, Sequence: 2, Status: PENDING, Output: VIDEO_002; VIDEO_003, Sequence: 3, Status: PENDING, Output: VIDEO_003. Continue until the final prompt. STEP 4 — CREATE THE MASTER FLOW AGENT: Create ONE main agent called MASTER VIDEO GENERATION ORCHESTRATOR. Its job is to control the entire bulk-generation process. The Master Agent must receive the complete VIDEO_PROMPT_MANIFEST. It must process every prompt sequentially. MASTER AGENT COMMAND: The Master Agent must follow this exact logic: START, Load the complete Video Prompt Manifest, Count the total number of video prompts, Set CURRENT_INDEX = 1 and TOTAL_VIDEOS = detected number of prompts, Select VIDEO_CURRENT, Send the EXACT ORIGINAL_PROMPT to the connected Text-to-Video generation tool, Wait for the generation request to complete, Confirm that a video output was successfully generated, Assign the output VIDEO_001, VIDEO_002, VIDEO_003, etc., Mark the current prompt STATUS = COMPLETED, Increase CURRENT_INDEX = CURRENT_INDEX + 1, Check: IF CURRENT_INDEX <= TOTAL_VIDEOS THEN generate the next video, IF CURRENT_INDEX > TOTAL_VIDEOS THEN finish the workflow. EXTREMELY IMPORTANT — SEQUENTIAL PROCESSING: The Agent must process prompts in exact order, for example VIDEO_001, VIDEO_002, VIDEO_003, VIDEO_004, VIDEO_005. Never randomly reorder them. Never prioritize one prompt over another. Never skip a prompt. Never generate VIDEO_005 before VIDEO_004. NO DUPLICATE GENERATION: The system must maintain a generation state. Each prompt can have only one of these statuses: PENDING, GENERATING, COMPLETED, FAILED, RETRYING. Before generating any prompt, check its status. If STATUS = COMPLETED, do NOT generate it again. If STATUS = PENDING, generate it. If STATUS = FAILED, follow the retry procedure. FAILURE / RETRY SYSTEM: If a Text-to-Video generation request fails: 1. Record the failure. 2. Keep the original prompt unchanged. 3. Retry the SAME prompt. 4. Maximum retries = 3. 5. Do not move to the next prompt until the current prompt succeeds or is explicitly marked for manual review. Retry flow: VIDEO_017, GENERATION FAILED, RETRY 1, FAILED, RETRY 2, FAILED, RETRY 3, FAILED, MARK FOR MANUAL REVIEW. Then continue to the next prompt only if the automation platform allows failed items to be skipped. Never silently delete a failed prompt. PROMPT INTEGRITY RULE: The Text-to-Video Agent must receive the original prompt exactly as provided. The automation layer is NOT allowed to modify the creative prompt. Do not paraphrase, shorten, translate, rewrite, summarize, improve, or reinterpret the prompt. The automation agent's job is ONLY to deliver the prompt to the video-generation tool. VIDEO OUTPUT NAMING: Automatically assign sequential output names: VIDEO_001, VIDEO_002, VIDEO_003, VIDEO_004 ... If the platform supports file naming, use Video_001, Video_002, Video_003. If metadata is supported, attach Project_ID, Video_ID, Sequence_Number, Original_Prompt_ID. CHARACTER / STORY CONSISTENCY: The automation agent must NOT modify the content of the prompts. However, it must preserve the original sequence. This is important because the prompts may represent consecutive scenes from one story. Therefore VIDEO_001 must remain first, VIDEO_002 must remain second, VIDEO_003 must remain third ... VIDEO_N must remain last. Never shuffle the order. Never merge scenes. Never regenerate an earlier scene because a later scene exists. BATCH PROCESSING: If the connected platform supports batch processing, use it. If batch processing is NOT supported, simulate batch processing using a sequential loop. The user should still only need to start ONE automation. The system must automatically process every prompt. The user must NOT be required to manually paste each prompt. RATE LIMIT / DELAY HANDLING: If the video-generation platform requires a waiting period between generations, automatically wait for the required interval. Do not terminate the workflow because of a normal generation delay. If the platform reports RATE_LIMIT, QUOTA_LIMIT, TEMPORARY_ERROR, SERVER_BUSY, or GENERATION_PENDING, handle the condition according to the platform's available retry/wait mechanism. Never lose the current prompt. GENERATION CONFIRMATION: After sending a prompt to the Text-to-Video tool, verify that the generation request was accepted. If the tool provides a Generation ID, Task ID, Video ID, Job ID, Output URL, or File ID, store it inside the current video's record. Example: VIDEO_023, Generation Status: COMPLETED, Generation ID: [ID], Output: Video_023. MASTER STATE TABLE: Maintain this state: Sequence 1, Video ID VIDEO_001, Prompt Original Prompt 1, Status PENDING, Attempts 0, Output —; Sequence 2, Video ID VIDEO_002, Prompt Original Prompt 2, Status PENDING, Attempts 0, Output —; Sequence 3, Video ID VIDEO_003, Prompt Original Prompt 3, Status PENDING, Attempts 0, Output —. Continue for every prompt. The table is an internal workflow state and must remain synchronized with actual generation status. MASTER ORCHESTRATOR VARIABLES: Create and use {{PROJECT_ID}}, {{TOTAL_VIDEOS}}, {{CURRENT_INDEX}}, {{CURRENT_VIDEO_ID}}, {{CURRENT_PROMPT}}, {{GENERATION_STATUS}}, {{RETRY_COUNT}}, {{GENERATION_ID}}, {{OUTPUT_VIDEO}}, {{COMPLETED_COUNT}}, {{FAILED_COUNT}}, {{REMAINING_COUNT}}. MAIN AUTOMATION LOOP: The actual logic must be: FOR EACH VIDEO PROMPT IN ORIGINAL ORDER: 1. Read current prompt. 2. Verify prompt exists. 3. Verify status. 4. Set status to GENERATING. 5. Send exact prompt to Text-to-Video tool. 6. Wait for generation. 7. Check generation result. 8. If successful, save output, mark COMPLETED, move to next prompt. 9. If failed, retry according to retry rules. 10. Continue until all prompts are processed. END LOOP. IMPORTANT — DO NOT STOP EARLY: If there are 60 prompts, process all 60. If there are 80 prompts, process all 80. If there are 100 prompts, process all 100. Do NOT stop after 10, 20, 30, or the first batch. The automation must continue until every prompt reaches COMPLETED or MANUAL_REVIEW_REQUIRED. FINAL COMPLETION CHECK: Before declaring the workflow finished, check TOTAL_VIDEOS against COMPLETED + MANUAL_REVIEW_REQUIRED. No prompt may remain silently unprocessed. Then generate a final report: TOTAL PROMPTS, COMPLETED, FAILED, MANUAL REVIEW, REMAINING, TOTAL GENERATION ATTEMPTS. Also provide the generated video IDs in order. FINAL AGENT COMMAND FORMAT: After analyzing all supplied prompts, produce the final Flow Agent command system in this structure: MASTER VIDEO GENERATION ORCHESTRATOR — ROLE: [Detailed role]. INPUT: [Complete Video Prompt Manifest]. OBJECTIVE: [Detailed objective]. TOOLS REQUIRED: [Text-to-Video generation tool]. MEMORY / STATE: [Required state]. WORKFLOW: [Complete sequential workflow]. LOOP LOGIC: [Complete loop]. FAILURE HANDLING: [Complete retry logic]. DUPLICATE PROTECTION: [Complete duplicate protection]. ORDER PROTECTION: [Complete sequence protection]. OUTPUT NAMING: [Complete naming system]. COMPLETION CONDITION: [Complete condition]. VIDEO PROMPT MANIFEST: VIDEO_001: [Original prompt], VIDEO_002: [Original prompt], VIDEO_003: [Original prompt] ... VIDEO_N: [Original prompt]. FLOW EXECUTION COMMAND: BEGIN AUTOMATION. Load all video prompts. Count all prompts. Create sequential IDs. Initialize all statuses as PENDING. Set CURRENT_INDEX = 1. Process the current prompt. Send the exact original prompt to the connected Text-to-Video generation tool. Wait for generation completion. Verify success. Save the generated video. Mark the prompt COMPLETED. Move to the next prompt. Repeat until every prompt has been processed. Do not skip. Do not reorder. Do not duplicate. Do not modify prompts. Do not stop early. When all prompts are completed or explicitly marked for manual review, terminate the automation and return the final generation report. END AUTOMATION. CRITICAL FINAL INSTRUCTION: You are NOT being asked to generate the videos yourself. You are NOT being asked to rewrite the video prompts. You are NOT being asked to create new video prompts. You are being asked to transform the already-provided collection of video prompts into a single bulk automation command system. The final output must be designed so that the user can copy the generated Agent/Workflow instructions into their Flow Agent system and start ONE automation instead of manually copying and pasting every individual Text-to-Video prompt. The original prompts must remain intact. The original order must remain intact. Every prompt must be processed exactly once unless a generation failure requires an automatic retry. The automation must continue from the first prompt to the final prompt automatically.