fix(core): clarify the read tool's offset validation error

[holny]

Issue for this PR

Fixes #46807

Type of change

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

What does this PR do?

Models occasionally hallucinate a negative value for the optional read offset (for example offset: -1), and the validation failure only said the value should be >= 0 without acknowledging that line and entry numbers are 1-based. Annotate the input schema so the failure carries the 1-based guidance and nudges the model toward omitting the offset or using 1.

Offset 0 remains valid and still normalizes to the first line, matching the existing zero-offset normalization behavior.

How did you verify your code works?

• New case in test/tool-read.test.ts: invoking read with offset: -1 surfaces the 1-based guidance
• bun typecheck in packages/core; bun test test/tool-read.test.ts (27 passing, including the zero-offset normalization case)

Checklist

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

---

Supersedes #47175 — the PR head fork was accidentally deleted on 2026-09-09 (not intentional, see the notification comment there); re-filing so the review can continue. Original conversation: #47175

[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:

Potential Related PR Found:

• PR refactor(core): localize read tool input schema #46810 - refactor(core): localize read tool input schema
  • refactor(core): localize read tool input schema #46810
  • This PR also modifies the read tool's input schema, which is directly related to the offset validation changes in PR fix(core): clarify the read tool's offset validation error #48160. While not a direct duplicate, it should be reviewed together since both address the read tool schema.

The search confirms that PR #48160 (the current PR) is a re-filing of #47175 (head fork was deleted), which is clearly stated in the description. No other duplicate PRs were found.