#376376th of 455 in Skills, plugins and prompts
agent-vision-toolkit
Vision CLIs and skill that let text-only coding agents read images
- Stars
- 1.3k
- License
- MIT
- Last commit
- Oct 2026
- Last release
- Aug 2026
- Language
- Python
4 on 28 Jul 20261,260 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Provides shell CLIs (glance, ground, detect, trace, crop) that send images to a multimodal API for Q&A, OCR, bounding boxes and SVG tracing. It also ships a skill with playbooks for long-screenshot OCR, UI restoration and GUI operation. An optional local proxy (Codex, Claude Code) or single-file extensions (Pi, Oh My Pi, OpenCode) make pasted images work with text-only models such as DeepSeek.
Who it is for: Users of text-only coding agents (e.g. DeepSeek) who need image input
Strengths
- Passes the agent's current intent to the vision model as a focus hint
- Works with any agent that can run shell commands, no integration required
- Accepts OpenAI Chat Completions, Responses or Anthropic Messages endpoints
- trace fits SVG primitives locally, without an LLM
Weaknesses
- Needs a separate multimodal API key, base URL and model for vision calls
- Descriptions default to Chinese; English requires setting LANG=en
- ground, detect, crop and trace need pillow; trace also needs numpy
- Proxy and extension integrations are verified only for five named agents
What it needs
- no GPU
- Needs Python 3.11+, multimodal vision API, pillow, numpy, vtracer (optional)
- Models: OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, DeepSeek (text-only host model)
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Skills, plugins and prompts
See all 455| Rank | Project | Score |
|---|---|---|
| 1 | claude-memPersistent cross-session memory plugin for Claude Code and other coding agents | 97 out of 100 |
| 2 | impeccableDesign skill with 24 slash commands for AI coding agents' frontend output | 97 out of 100 |
| 3 | gstackMarkdown slash-command skills that give Claude Code planning, review, QA and ship roles | 96 out of 100 |
| #4 | rufloAgent harness adding swarms, memory and hooks to Claude Code and Codex | 96 out of 100 |
| #5 | cavemanShortens coding-agent replies and compresses tool output before the model reads it | 95 out of 100 |