fix(opencode): fix the read tool reports image/pdf read successfully when the LLM lacks vision capabilities

[Demonese]

Issue for this PR

Closes #21800

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

The description of the read tool can mislead text-only LLMs, causing them to hallucinate that they can understand image content.

To fix this issue, I first inspected the source code of the read tool and found that it does not verify the LLM vision capabilities when handling images and PDF files.

So fixed it.

How did you verify your code works?

See Screenshots / recordings

Screenshots / recordings

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found one potentially related PR that should be reviewed:

Related PR:

• feat(tool): Add media attachments to read tool #22258 - feat(tool): Add media attachments to read tool
  • This PR adds media attachment support to the read tool, which is closely related to the vision capabilities issue addressed in PR fix(opencode): fix the read tool reports image/pdf read successfully when the LLM lacks vision capabilities #22263. Since both PRs deal with how the read tool handles images/media, there could be overlap or conflicts that need to be resolved.

The other results (#21917, #11333, #15262) are related to PDF/image handling but appear to be older or more specific fixes that may not directly conflict with the current PR.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍