feat(cli): read Jupyter notebooks as cell content

[marius-kilocode]

Reading .ipynb files currently sends raw notebook JSON into model context, including metadata and output payloads that obscure the useful document content.

This extracts markdown and code cell sources in document order for valid notebooks while preserving the read tool's normal pagination and truncation behavior. Malformed notebooks continue through raw-text reading so they remain inspectable, and native PDF and image attachment handling stays unchanged.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Previously Reported — Now Fixed

File	Issue
packages/opencode/src/kilocode/tool/notebook.ts	Inconsistent bail-out vs. skip strategy for malformed cells ✅
packages/opencode/src/kilocode/tool/notebook.ts	Double file read for malformed .ipynb files ✅
packages/opencode/src/kilocode/tool/notebook.ts	Empty-cells edge case — cells.join("\n\n") returned "" when all cells were skipped ✅

Files Reviewed (4 files)

• packages/opencode/src/kilocode/tool/notebook.ts — clean (empty-cells fix applied)
• packages/opencode/test/kilocode/read-notebook.test.ts — clean (new test covers the fixed edge case)
• packages/opencode/src/tool/read.ts — unchanged, clean
• .changeset/read-notebook-cells.md — unchanged, clean

---

_{Reviewed by claude-sonnet-4.6 · 164,911 tokens}

_{Review guidance: REVIEW.md from base branch main}