[Bug] look_at tool fails with image input even when modalities configured

[Ryker0716]

Bug Description

The look_at tool fails to analyze image files, returning error "Cannot read image (this model does not support image input)" even when the model has modalities configured for image input support.

Steps to Reproduce

1. Configure custom provider with modalities in ~/.config/opencode/opencode.jsonc:

{
  "provider": {
    "rundao": {
      "models": {
        "public/qwen3.5-397b": {
          "modalities": {
            "input": ["text", "image"],
            "output": ["text"]
          },
          "capabilities": {
            "input": {
              "image": true
            }
          }
        }
      }
    }
  }
}

2. Try to use look_at tool with an image file

3. Observe error: "Cannot read image (this model does not support image input)"

Expected Behavior

The look_at tool should respect the modalities configuration and pass image files to vision-capable models.

Actual Behavior

• look_at tool fails with "model does not support image input" error
• After removing oh-my-opencode plugin, image input works correctly
• Direct image paste/drop into chat works fine

Root Cause

oh-my-opencode's tool handling layer doesn't properly pass the modalities capability check to the look_at tool.

Workaround

Remove oh-my-opencode plugin from config:

{
  // Remove this line:
  // "plugin": ["oh-my-opencode@latest"],
  "$schema": "https://opencode.ai/config.json",
  ...
}

After removing the plugin, image input works correctly.

Environment

• OpenCode version: latest
• oh-my-opencode version: latest
• OS: macOS
• Model: qwen3.5-397b (vision-capable)

Related

• OpenCode #15728
• OpenCode #15040

[372572571]

Maybe you need to use look_at("resource path method") to clarify; this issue seems to have existed for a long time. I've been using it this way to bypass this bug.