How do I install the Vision skill?

Run npx skillmds@latest add lqc2007224-max/vision in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Vision safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: executes scripts, makes network calls, reads secrets. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Vision?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Vision free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Vision?

lqc2007224-max (@lqc2007224-max) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Vision by lqc2007224-max — Agent Skill

Name: Vision
Rating: 4.2 (5 reviews)
Author: lqc2007224-max

File contents

Vision Skill — External Multimodal Image Analysis

This skill uses Gemini 2.5 Flash (free tier, 250 req/day) via Google AI Studio API to analyze images when the primary model cannot process visual content natively.

When to Use

Invoke this skill whenever the user:

Provides an image path and asks you to describe, analyze, or extract information from it
Asks "what's in this image/screenshot/photo?"
Needs text extracted from a screenshot or UI
Wants to understand a diagram, chart, or design mockup
Appends an image file and expects visual understanding

How to Use

Run the helper script, then parse the JSON response:

bash .claude/skills/vision/vision.sh "<image_path>" "<prompt>" [model]

Parameters

image_path (required): Absolute or relative path to the image file
prompt (optional): What to ask about the image. Be specific! Default is a general description.
model (optional): Defaults to gemini-2.5-flash. Also available: gemini-2.5-pro, gemini-3.1-flash-lite

Environment

GEMINI_API_KEY: Your Google Gemini API key (get one free at https://aistudio.google.com/apikey)
HTTPS_PROXY: Falls back to http://127.0.0.1:7897 (Clash Verge)

Parsing the Response

The script outputs raw JSON. Extract the text response from:

.candidates[0].content.parts[0].text

For errors, check:

.error.message

Fallback — When Gemini Quota is Exhausted

If the Gemini API returns HTTP 429 or an error message containing RESOURCE_EXHAUSTED, the daily free quota (250 req/day) has been used up. In this case, automatically fall back to the backup vision skill:

bash .claude/skills/vision-backup/vision-backup.sh "<image_path>" "<prompt>"

The backup skill uses Qwen-VL-Max via DashScope API. It accepts the same parameters and returns a similar JSON response. Parse the result from .choices[0].message.content.

Important: Always try Gemini first, only use the backup when Gemini is unavailable or quota-exhausted.

Best Practices

Be specific in your prompt — "What error dialog is shown in this screenshot?" works better than "Describe this."
For UI screenshots — Ask about layout, text content, buttons, error states, or specific elements.
For diagrams/charts — Ask about data, trends, or specific labeled elements.
For text extraction — Say "Extract ALL text from this image, preserving the layout."
Embedded images in conversation — If the user references an image file in their message, read it using the Read tool first to confirm it exists, then pass the path to the script.

Supported Formats

PNG, JPEG/JPG, GIF, WebP, BMP, SVG

---name: visiondescription: Analyze images using Google Gemini vision model when the current model lacks multimodal capabilities. Invoke with an image path to get AI-powered image descriptions, text extraction, or visual analysis.---# Vision Skill — External Multimodal Image AnalysisThis skill uses **Gemini 2.5 Flash** (free tier, 250 req/day) via Google AI Studio API to analyze images when the primary model cannot process visual content natively.## When to UseInvoke this skill whenever the user:- Provides an image path and asks you to describe, analyze, or extract information from it- Asks "what's in this image/screenshot/photo?"- Needs text extracted from a screenshot or UI- Wants to understand a diagram, chart, or design mockup- Appends an image file and expects visual understanding## How to UseRun the helper script, then parse the JSON response:```bashbash .claude/skills/vision/vision.sh "<image_path>" "<prompt>" [model]```### Parameters- `image_path` (required): Absolute or relative path to the image file- `prompt` (optional): What to ask about the image. Be specific! Default is a general description.- `model` (optional): Defaults to `gemini-2.5-flash`. Also available: `gemini-2.5-pro`, `gemini-3.1-flash-lite`### Environment- `GEMINI_API_KEY`: Your Google Gemini API key (get one free at https://aistudio.google.com/apikey)- `HTTPS_PROXY`: Falls back to `http://127.0.0.1:7897` (Clash Verge)## Parsing the ResponseThe script outputs raw JSON. Extract the text response from:```.candidates[0].content.parts[0].text```For errors, check:```.error.message```## Fallback — When Gemini Quota is ExhaustedIf the Gemini API returns **HTTP 429** or an error message containing `RESOURCE_EXHAUSTED`, the daily free quota (250 req/day) has been used up. In this case, **automatically fall back** to the backup vision skill:```bashbash .claude/skills/vision-backup/vision-backup.sh "<image_path>" "<prompt>"```The backup skill uses **Qwen-VL-Max** via DashScope API. It accepts the same parameters and returns a similar JSON response. Parse the result from `.choices[0].message.content`.**Important:** Always try Gemini first, only use the backup when Gemini is unavailable or quota-exhausted.## Best Practices1. **Be specific in your prompt** — "What error dialog is shown in this screenshot?" works better than "Describe this."2. **For UI screenshots** — Ask about layout, text content, buttons, error states, or specific elements.3. **For diagrams/charts** — Ask about data, trends, or specific labeled elements.4. **For text extraction** — Say "Extract ALL text from this image, preserving the layout."5. **Embedded images in conversation** — If the user references an image file in their message, read it using the Read tool first to confirm it exists, then pass the path to the script.## Supported FormatsPNG, JPEG/JPG, GIF, WebP, BMP, SVG

Frequently asked questions

Vision