@op7418

youtube-clipper downloads youtube videos and subtitles, uses ai to generate fine-grained chapters (2-5 min), lets users select segments for automated clipping, bilingual subtitle translation, subtitle burning, and summary generation. strongest at procedural detail, weakest on trigger clarity and recovery paths.

logo-generator collects brand metadata through conversational intake and promises svg and presentation output, but lacks concrete generation logic, output guarantees, or failure handling.

claude-to-im bridges claude chat sessions to telegram, discord, feishu/lark, qq, and wechat. skill configuration parsing is incomplete and lacks procedural clarity, failure handling, and functional specification.

guizang-ppt-skill generates single-file HTML presentations in two aesthetic modes: magazine-style with serif typography and swiss grid-based design. strong operational workflows for requirements clarification, template selection, and content filling, but lacks concrete trigger phrases and explicit failure mode handling.

document-illustrator generates styled illustrations from document content by extracting key topics and rendering them in one of three visual styles. lacks concrete procedural steps and operational guardrails.

humanizer-zh detects and removes ai writing patterns from text. scans 20+ common markers including overused punctuation, vague phrasing, and filler words, then rewrites sections to sound more natural while preserving meaning and tone.

Generate Guizang-style social card image sets, Live Photo motion cards, material-first Live Photo puzzle layouts, triple Live Photo collages,…