Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/Lingtai-AI/lingtainpx agentmods add skills/lingtai-ai/lingtai/visionWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/lingtai-ai/lingtai/vision)<a href="https://agentmods.dev/skills/lingtai-ai/lingtai/vision"><img src="https://agentmods.dev/badge/skills/lingtai-ai/lingtai/vision/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/lingtai-ai/lingtai/vision"><img src="https://agentmods.dev/badge/skills/lingtai-ai/lingtai/vision.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00120 | $0.01064 |
| Opus 5 | $0.00060 | $0.00532 |
| Sonnet 5 | $0.00024 | $0.00213 |
| Haiku 4.5 | $0.00012 | $0.00106 |
Grade A, and why
vision scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
vision (router)
Nested swiss-knife reference for image understanding. Three paths — pick the cheapest available.
Decision Tree
Is `vision` tool in your tool list?
├── YES → use it directly, done.
└── NO → does the user have a MiniMax preset/key slot?
(read the sibling `../minimax-cli/SKILL.md`; scan ~/.lingtai-tui/presets/** without printing keys)
├── YES → use the `minimax-cli` reference for `mmx vision …` from the shell, or an already-registered MCP tool
└── NO → fall back to local VLM:
python3 <skill-path>/scripts/describe.py <image>
See reference/local-models.md.
Path 1 — Built-in vision Tool
If your LLM provider supports image input (MiniMax, Gemini, Anthropic, OpenAI, Zhipu), the kernel exposes a vision tool directly. Cheapest, lowest latency, no extra setup.
vision not in your tool list? Your LLM provider doesn't support image input — fall through to Path 2 or 3.
Path 2 — MiniMax via the sibling minimax-cli reference
For text-only LLMs (DeepSeek, OpenRouter text-only, Codex) with a usable MiniMax preset/key slot. Two routes can exist:
- Shell —
mmx vision …via the official CLI. No MCP registration needed; just install + key. Best for ad-hoc one-shots in bash. - In-tool — an already-registered MiniMax MCP vision tool. Best when the agent needs vision as a tool call inside a longer reasoning loop. MCP server registration is owned by
mcp-manual(kernelmcpcapability).
Read the sibling minimax-cli reference (../minimax-cli/SKILL.md) before using the shell route. It owns the canonical MiniMax credential discovery flow: scan TUI presets recursively, pick the declared slot, export it without printing the key, and match the preset region/base URL. Do not assume a bare MINIMAX_API_KEY is the only valid slot.
Path 3 — Local VLM (offline, unlimited)
For agents that need image analysis without a vision-capable LLM and without a MiniMax key. Or for batch jobs / privacy-sensitive content / unlimited quota.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 74 lines · 120 tokens per session scan A 4965fa959d14
vision is a skill published in the GitHub repository Lingtai-AI/lingtai (670 stars, last pushed today), licensed Apache-2.0. It adds 120 tokens to every session and 1,064 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
multimodal-llm
Vision, audio, video generation, and multimodal LLM integration patterns. Use when processing images, transcribing audio, generating speech, generating AI video (Kling v3, Sora 2, Veo 3.1 std/lite/fast, Runway Gen-4.5 via gen4turbo), or building multimodal AI pipelines.
tao-generate-referring-expressions
Four-step image referring-expression pipeline: turns images plus KITTI bounding-box labels into region descriptions, scene captions, grounded referring expressions, and (optionally) verified expressions via VLM distillation. Use when the user wants to generate referring-expression annotations from images with KITTI…
tao-generate-image-grounding
Two-step image grounding pipeline: extracts referring expressions from (image, caption) pairs and grounds them to pixel-space bounding boxes via a VLM. Use when the user wants to ground captions to bboxes, generate phrase-grounded annotations, auto-label images for grounding, or run the imagegrounding pipeline.…
glmv-prompt-gen
Analyze images/videos and generate professional prompts for text-to-image and text-to-video AI tools (Midjourney, Stable Diffusion, DALL-E, Sora, Runway, Kling, Pika). Use when the user wants to generate prompts from reference images/videos, create AI art prompts, or get prompt engineering suggestions from visual…
glmv-resume-screen
Screen and evaluate resumes against criteria using ZhiPu GLM-V multimodal model. Reads multiple resume files (PDF/DOCX/TXT), compares against user-defined screening criteria, and outputs a Markdown table with pass/fail analysis. Use when the user wants to filter resumes, compare candidates, or batch-evaluate job…
image-understanding
Enable image understanding capabilities for describing/analyzing images, reading text, and answering questions about visual content.