LongLeo
Community@LongLeo287 · Ho Chi Minh City
LongLeo's SEOSONA skill suite governs AI image/video generation, prompt engineering, quality-gating, and local-first media pipelines for Vietnamese content creators.
Agent Skills by LongLeo
Showing 15 vetted skills indexed across 1 GitHub repositories.
LongLeo287flow-mcp
Generates images and videos through SEOSONA Flow via a local MCP server.
LongLeo287build-capability
Guides building local-first CLI tools, DSL parsers, search indexes, and storage layers from scratch.
LongLeo287mllm-judge
Scores AI-generated images and videos against rubric criteria using vision models as judges.
LongLeo287img-to-3d
Converts a reference image into a procedural Three.js model via code-only primitives.
LongLeo287video-ai-v2-pipeline
Plans local video production pipelines with TTS, ASR, captioning, and reframing via sidecar tools.
LongLeo287prompt-optimize
Refines image and video generation prompts through an optimize-evaluate-iterate loop.
LongLeo287interview-to-spec
Interviews users to complete image and video generation briefs before writing prompts.
LongLeo287video-prompt-scaffold
Generates structured seven-part video prompts for Veo, Sora, and Kling generation models.
LongLeo287viral-ad-formats
Provides Vietnamese UGC ad video format templates and product image presets for Veo generation.
LongLeo287spec-gate-build
Validates JSON specs before building structured artifacts in deterministic passes with verification.
LongLeo287text-in-image
Generates correctly spelled text in AI images via reserve-and-overlay or constrained bake-in prompting.
LongLeo287anti-slop-visual
Lint AI image and video generation prompts against banned filler phrases and enforce concrete descriptive terms.
LongLeo287image-qa
Reviews AI-generated images against a visual checklist and OCR text comparison to issue accept, regenerate, or overlay verdicts.
LongLeo287character-consistency
Maintains character and style consistency across batch AI image generation prompts.
LongLeo287video-production
Plans multi-scene AI videos with voiceover-timed beats, style-anchor keyframes, and per-beat resumable state.
Frequently Asked Questions About LongLeo
FAQPage SchemaWhat tasks can I accomplish with LongLeo's SEOSONA skills?▼
You can generate and refine AI images and videos via SEOSONA Flow, optimize prompts for Veo/Sora/Kling, enforce character consistency across scenes, QA images for text and anatomy defects, build procedural Three.js models from reference photos, and plan local TTS/caption/reframe video pipelines.
Who are these skills designed for?▼
They target Vietnamese content creators, video ad producers, and developers building the SEOSONA Flow extension and Video AI V2 product. Users range from marketers needing viral UGC-style b-roll to engineers implementing spec-gate-build architectures and local sidecar media binaries.
How do the generation and quality-control skills work together in practice?▼
A typical flow: interview-to-spec clarifies the brief, video-prompt-scaffold writes disciplined prompts, anti-slop-visual bans vague tokens, Flow generates assets, then image-qa and mllm-judge score outputs with accept/reject verdicts. Video-production orchestrates multi-scene results with resumable per-beat state.
Do these skills require paid APIs or external services?▼
The viral-ad-formats taxonomy explicitly uses no paid APIs—Flow generates b-roll with your own Veo access. The MCP server is optional, and Video AI V2 heavy components (whisperX, audio.cpp) build in separate sidecar repositories, keeping the core extension lightweight and local-first.
What are the prerequisites for using the video pipeline skills?▼
Video AI V2 pipeline skills are plan/interface definitions only; actual binaries like the C++ audio.cpp and Python whisperX must be built in separate sidecar repos. The flow-mcp skill requires a local MCP server, though both Flow and V2 run independently without it.