AI Agent Workflow Collection

Use when a task involves Feishu/Lark Minutes, Docx intelligent meeting notes, transcript documents, or Feishu source-routing for Chinese interview review, coaching, or preparation reports. Pasted transcripts are supported only as a fallback when Feishu transcript sources are unavailable.

Create animated GIF showcases from multiple input images. Use when the user provides images and wants an ordered GIF demo, image sequence animation, feature showcase, product preview, screenshot carousel, or before/after presentation. Defaults to upload/input order, hold frames per image, fade transitions, automatic canvas sizing, and a local GIF output path.

Use when the user provides screenshots or images and wants them OCRed, converted, continued, or written into a specified Markdown file, especially for notes with headings, tables, formulas, emoji, highlighted text, or multi-batch image input.

Review job-related contracts or pasted contract text for red flags, warnings, protective clauses, missing protections, fairness score, and negotiation suggestions. Use when the user provides PDF, DOCX, TXT, MD, RTF, or pasted text for employment agreements, offer letters, NDAs, contractor agreements, freelance agreements, internship agreements, non-competes, severance terms, or asks for practical contract risk review.

Capture project-scoped learning notes, technologies, concepts, glossary terms, architecture decisions, implementation lessons, and review questions into a durable Markdown learning log. Use when the user says to remember, record, log, add to review notes, add to the learning doc, save this concept, keep this term for later, or update {{projectname}}学习复盘技术栈与概念记录.md while working in a repository.

Use OpenSense to find, evaluate, and prepare one safe open-source PR attempt from watched GitHub repositories. Use when the user wants daily issue recommendations, LLM-assisted issue planning, context packs, sandboxed agent apply, test evidence capture, PR draft generation, attempt status, or read-only OpenSense MCP/Skill integration.

Use when the user provides a research paper screenshot, title, arXiv/OpenReview/publisher/project URL, or article excerpt and wants Codex to identify the paper, download and verify an official PDF, classify it into the 01.raw/08.Research taxonomy, choose an optional filename prefix with --name-prefix, and report a Zotero Add Item by Identifier value such as arXiv ID or DOI.

Use when creating or updating VectorPeak's personal GitHub profile README with Chinese-first/English-second bilingual sections, badges, identity line, own public project summary, own merged upstream PR summary, stars-sorted personal Projects table, stars-sorted Open Source Contributions tables, and LinkedIn link.

Explain project-specific concepts, technical terms, models, algorithms, configuration items, dataset fields, architecture decisions, and codebase terminology while working in a repository. Use when the user asks to explain a term, understand a project, generate a project knowledge map, build a glossary, compare related concepts, clarify technologies in README/docs/scripts/configs/notebooks/reports, or study key project ideas top-down.

Deep rewrite authorized technical source corpora into polished Markdown blog posts using a user-provided topic and template. Use when the user provides source-corpus Markdown, WeChat/Zhihu URLs, screenshots, OCR text, or mixed materials and wants a logic-first technical article, iterative rewrite, blog-style draft, template-filled Markdown, or source-backed technical documentation with references.

Clip arbitrary WeChat Official Account articles through TikHub into Obsidian Web Clipper-style Markdown bundles. Use when the user provides a WeChat public account author/name, mp.weixin.qq.com article URL, article title, screenshot/OCR text, copied page/search text, or asks to fetch selected WeChat public-account articles and save Markdown clippings into a local clippings directory or Obsidian vault folder.

Clip Zhihu content into Obsidian Web Clipper-style Markdown bundles. Use when the user provides a Zhihu person name or people URL, article URL/title, question URL, answer URL, screenshot/OCR text, copied snippet, or mixed clues and wants official-first identity/search verification, TikHub-final fulltext retrieval, answer clipping, grouped Markdown output, or raw Zhihu clippings saved to a local inbox/vault folder.