热门工作流

With just one sentence of inspiration, you can generate: short drama/short video script (or screenplay), storyboard script, JSON production package, and automatically generate keyframe images and finished video (mp4) with the help of the Winning Cloud video model.

Upload a UTF-8 text or Markdown PRD and receive an evidence-grounded four-role red-team review before development begins. The report covers Product, Engineering, QA, and Business/Operations; returns a PRD Quality Score and PASS / REVIEW / BLOCK decision; and highlights top must-fix issues, missing requirements, missing acceptance criteria, edge cases, contradictions, AI Coding Readiness, and recommended revisions. Upload UTF-8 text or Markdown PRD to get evidence-based red team reviews for product, engineering, QA, and business/operations before development begins. The report includes PRD quality score, PASS / REVIEW / BLOCK, top fix items, missing requirements, missing acceptance criteria, boundary and abnormal scenarios, conflicts, AI coding readiness and modification suggestions. Passed representative private LoomLoom smoke validation: 10/10 cases, 95/95 grounded findings, 0 unsupported factual claims, 3/3 expected Critical recalls, and 0 prompt-injection successes. The 60-case hosted runtime evaluation was not completed. Passed the representative LoomLoom private smoke verification: 10/10 use cases passed, 95/95 found evidences were implemented, 0 unfounded facts, 3/3 expected critical issue recalls, 0 prompt injections were successful. 60 hosting run assessments not completed. This SkillBot provides AI-assisted PRD review and development-readiness analysis. Final product, engineering, security, compliance, and launch decisions should remain under human review. This SkillBot provides AI-assisted PRD review and development readiness analysis. Final product, engineering, safety, compliance and release decisions should still be reviewed manually. Current limitations: the completed hosted smoke used English output; PDF/DOCX files require text conversion; model/API cost is billed separately and varies with usage. Current limitations: Completed hosted smoke uses English output; PDF/DOCX needs to be converted to text first; model/API costs are billed separately and vary with actual usage.

One-sentence inspiration → Complete novel (v3 improved version): Based on v2, fix naming consistency, ban AI response language, hard constraints on delivery scale, and remove formatting noise. Built-in 24 style matrix + audience rhythm curve + AI-free blacklist, automatically generates book title, world view, characters, plot, main text (full text of the first 3 chapters of the standard package) and bookstore introduction. The buyer fills in the inspiration, audience/style and delivery scale, and gets a set of novel materials that can be published directly in one run.

Huan Ling himself is the author of "The Boundary of AI Soul". Not a universal AI assistant, but the voice of this book. He only talked about the propositions in the book and kept silent about the rest. Support reader interviews/chapter discussions/AI and humanities dialogue. Starting from June 10, 2026, Xiaohongshu Daily will serialize 31 chapters from D0 to D30. Enter what you want to say to Huan Ling or questions about this book, and Huan Ling will respond as the author.

Batch evidence-based resume screening for recruiter review. Extracts job-relevant evidence, assesses explicit requirements, audits unsupported claims and protected-attribute leakage, and returns traceable recommendations. PDF resumes must be parsed to text before submission.

Generate optional scripts and video material search terms in batches based on video themes, and support quotation first and execution after user confirmation.

Generate short video materials without text and watermark for MoneyPrinterTurbo; execute them in batches based on scenes and return MP4 artifacts.

The user provides image prompts for 4 scenes and generates 4 16:9 scene pictures (departure/dilemma/enlightenment/delivery), Low-poly × Pixel Art × Island Diorama style, no characters, pure scene narrative. v3 upgrade: supports frame-by-frame independent description.

The user inputs the HTML source code, and then completes the AI-flavored optimization of the HTML source code according to the disassembled hallmark requirements and steps.

Batch portfolio visual evaluation: Use the Gemini visual model to read screenshots of designer portfolio pages, evaluate the quality of visual design and depth of interactive cases, and output scores, highlights, risks, and interview suggestions. Used in conjunction with cv-hunter, it forms a complete recruitment evaluation system of L1 resume + L2 portfolio.

Batch resume and job matching screening: Each resume outputs matching scores (0-100), stratification (recommended/pending/elimination), scoring evidence with resume original text citations, missing items and job candidate rankings for all positions. Suitable for HR and headhunters to quickly identify target candidates among a large number of resumes. Billed by task.

AI University Course Selection and Scheduling Assistant: Enter the school, major, grade and your goals, and it will produce three results in one execution - in-depth analysis of the course catalog, course selection recommendations customized according to goals (with reasons and risk warnings for each course), and a semester course scheduling plan with three-color priority of must-take/recommended/alternative. Supports batch processing of multiple universities/students in one run (Excel workbook, one per row). [Important: Data source description] This SkillBot is a pure text processing pipeline and is not connected to the Internet. It needs to provide course catalog text before it can produce valuable content: · Recommended usage - let any AI Agent (WorkBuddy / Codex / Claude Code, etc.) first check the training plan/course manual on your school's official website online, organize it into a text and give it to you, and then submit it to the cloud. The agent is responsible for "finding data", and the pipeline is responsible for "processing data in batches, stably, and in a unified format." · Manual usage - paste the training plan text from the school's official website into the Course Catalog column of the workbook (or upload PDF/TXT). [Complete product includes local support] The SkillBot installation package comes with a complete local Skill (including Excel 6/7-sheet course schedule, ICS calendar export, conflict detection/anti-hallucination verification, etc.), see GitHub ez-hq/uni-course-scheduler. Cost: ¥0.5 per execution + platform model cost (settled by token). The estimated fee is displayed before execution and will be deducted after confirmation.

Upload a product photo on a white background and enter product details to generate eight Amazon-style images: main image, brand value, key visual, selling points, lifestyle scene, feature benefits, details and dimensions.

Let any SkillBot creator generate promotional materials with one click: 1 4:5 main poster + 4 sets of platform copy + Chinese and English bilingual + automatically generated self-contained HTML page (including Chinese and English tab switching + copy button + LoomLoom trial CTA). Can be sent directly to Xiaohongshu/Moments/community/general channels.

New user activation diagnosis and optimization Focusing on a clear definition of activation (a new user completes a core behavior within N days after registration), perform data quality inspection, bottleneck location, time trend and group diagnosis, cause evidence grading and experimental design on standardized funnel data, and output an executable activation optimization decision brief. Suitable for businesses such as SaaS, App, e-commerce, content products, communities and API platforms. The final output is a complete Markdown diagnostic report, covering seven sections: data quality, funnel evidence, time diagnosis, group diagnosis, cause hypothesis, experimental design, and decision briefing. Upload data format (CSV): long table structure, columns = period, segment_type, segment_value, stage_order, stage_name, user_count - period: observation period (such as 2026-01 / 2026-W27) - segment_type: all or segmentation dimension (such as channel / platform) - segment_value: fill in all when all; fill in specific value when dividing into groups (such as official website / iOS) - stage_order: funnel step sequence number (starts from 1, must be monotonic) - stage_name: Funnel step name (such as completed registration / first effective behavior) - user_count: the number of users in this cycle-group-step Contains at least 2 periods; contains 1 row of segment_type=all, segment_value=all for the overall summary (allowed to be equal to the sum of each segment); if segment diagnosis is required, please add the corresponding row.

Upload the hand-drawn house floor plan, select the decoration style (medieval style/modern style/American/Japanese style/Nordic/minimalist/light luxury/industrial style), fill in the residents and functional requirements, and automatically generate a complete interior design plan in 5 steps: house type identification → standardized CAD drawing description → furniture layout and circulation planning → text description of space renderings → final Markdown plan document. Covers wall size tables, door and window details, storage system planning, material color matching and soft decoration list, and construction precautions. It is suitable for interior designers to quickly come up with plans and owners to provide decoration reference.

Research each U.S. stock using a value-investing framework: business model, competitive moat and capital allocation, financial quality, valuation, and margin of safety. The cloud workflow prepares three-statement analysis, a DCF, an evidence audit, and a report. For research assistance only; not investment advice. It provides no price targets, return forecasts, or buy/sell recommendations.

Enter an article, and the WeChat official account summary, Xiaohongshu copy, and three image prompts will be generated at once, and summarized into a multi-channel publishing plan that can be directly executed. One piece of content, efficiently published on multiple platforms. This workflow will first analyze the article's topic, core ideas, target readers and communication highlights, and then generate corresponding content for different channels: - Extract article summaries suitable for WeChat official accounts - Generate Xiaohongshu copy with title, body text and hashtags - Output 3 complete text prompts with pictures - Summarize content positioning, release focus, illustration suggestions and release order - Reduce repeated rewriting and quickly adapt an article to multiple content platforms Suitable for: content creators, self-media operators, brand marketers, local businesses, new media teams and individual entrepreneurs. Input content: article title and article body. Output content: article analysis, WeChat official account summary, Xiaohongshu copywriting, 3 image prompts and final release plan. Author: Angel

Package the creator's existing skill (natural language description or local Agent Skill) into LoomLoom TemplateSpec with one click, allowing you to list it on the Market and start collecting pay-per-view. No need to understand TemplateSpec JSON. 4-step automatic process: ① Refining skill intent ② Generating TemplateSpec ③ Local contract verification ④ Output "SkillBot Creator Guide". Suitable for: Creators who want to productize their professional capabilities but don’t want to learn the TemplateSpec contract.