WALKERKILLERWALKERKILLER/Look-Tongji-Notes/vision-support/AGENTS.md

Look-Tongji-Notes / vision-support

AGENTS.md53 starsChanged 5 months agoMITView on GitHub

Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llmwiki/AGENTS.mdAGENTS.md
- look-tongji-notesSkill
- addSkill
- cheatsheetSkill
- listSkill
- noteSkill
- pageSkill
- ralphtransSkill
- setupSkill
- transSkill
- wikiSkill
- vision-supportSkill

The file

We split this AGENTS.md into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
After modifying scripts/vision.mjs, test with node scripts/vision.mjs --help for argument parsing
line 36

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

vision-support

Purpose

Image recognition bridge skill for non-multimodal LLMs (e.g., DeepSeek, GLM). When the main model cannot process images, this skill delegates vision requests to a configured external vision model (OpenAI, Gemini, Qwen-VL, etc.), supports multi-image analysis, automatic fallback across multiple models, and 19+ API platforms. Includes an interactive init setup wizard and a comprehensive config management CLI.

Key Files

File	Description
`README.md`	English documentation with installation, usage, and configuration guide
`README.zh.md`	Chinese documentation (same content as README.md)
`SKILL.md`	Agent skill entry point: defines when/how the vision skill auto-triggers in agents
`package.json`	npm package definition with bin entry (`vision-support`), postinstall hook, and metadata
`config.example.json`	Configuration template with example model definitions (OpenAI, Gemini, Qwen-VL, GLM-4V) and proxy settings
`install.mjs`	Cross-platform Node.js installation script with interactive agent selection and uninstall support
`install.sh`	Mac/Linux/WSL shell installation script with git-based cloning

Subdirectories

Directory	Purpose
`bin/`	npm CLI entry points: `cli.mjs` (global `vision-support` command) and `postinstall.mjs` (npm postinstall hook that auto-deploys skill files to known agent directories)
`scripts/`	Core vision logic: `vision.mjs` -- a zero-dependency Node.js script handling image loading, multi-provider API calls (openai/google/anthropic formats), automatic proxy detection, model fallback, and interactive config management
`references/`	Reference documentation: `supported-models.md` -- lists tested vision-capable models, provider types, API formats, and recommended configurations (budget, premium, local-only, Chinese providers)
`images/`	Screenshots used in README: `usage-recognize.png` (terminal running vision-support) and `usage-result.png` (recognition result output)

For AI Agents

Working In This Directory

This directory is a standalone skill package cloned from https://github.com/penfick/skills. The core logic lives in scripts/vision.mjs (1300+ lines, zero external dependencies using only Node.js built-in modules). The SKILL.md defines when agents auto-trigger the skill (image attached, user mentions "screenshot"/"picture", manual /vision-support command). Configuration follows the iron rule: models configured here are used ONLY for image recognition and never participate in main logic reasoning.

Testing Requirements

After modifying scripts/vision.mjs, test with node scripts/vision.mjs --help for argument parsing
Config commands can be tested: node scripts/vision.mjs config list (requires a config.json)
Image recognition tests require a valid API key for at least one configured model
The install scripts (install.mjs, install.sh) should be tested in a clean environment

Common Patterns

Zero-dependency pattern: all HTTP requests, image processing, and CLI interaction use Node.js built-in modules (fetch, fs, path, readline, crypto)
Provider abstraction: callModel() dispatches to callOpenAI(), callGoogle(), or callAnthropic() based on the provider field
Proxy auto-detection: environment variables -> Windows registry -> port probing, with CONNECT tunnel patching for HTTPS
Model fallback chain: primary model first, then ordered fallbacks; VISION_DEFAULT_MODEL env var overrides the primary
Configuration is synced across all installed skill locations when saved

Dependencies

Internal

Referenced from ../repo/Look-Tongji-Notes/ as a supporting skill for slide image processing in the Tongji Look lecture pipeline

External

Node.js 18+ runtime only -- zero npm dependencies
External API services: OpenAI, Google Gemini, Anthropic Claude, and other vision model providers (19+ platforms)
Proxy detection: auto-detects Clash, V2Ray, Shadowsocks, and other local proxy ports

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free