Improve tool result rendering for multiple text items

[jancurn]

Summary

This PR enhances the rendering of tool call results by:

1. Renaming extractSingleTextContent to extractAllTextContent to better reflect its new behavior
2. Extending the function to handle multiple text content items (joining them with newlines) instead of only single items
3. Hiding the _meta field in human-mode output to reduce noise while keeping it available in --json mode
4. Improving documentation and test coverage

Key Changes

• Function rename and behavior change: extractSingleTextContent() → extractAllTextContent()

  • Now returns all text content joined with newlines when content is an array of only type: "text" items
  • Previously only worked with single text items
  • Returns undefined if content mixes text and non-text items (stricter validation)

• Output filtering: Tool result rendering now strips the _meta field in human mode

  • Reduces protocol noise in CLI output
  • Full payload remains available via --json flag
  • Applied in renderCallToolResult() for both tools-call and tasks-result commands

• Documentation improvements:

  • Updated JSDoc comments to clarify the new behavior
  • Added note that structuredContent is skipped when text content is rendered
  • Updated comments in output.ts and tools.ts to explain the rationale

• Test updates:

  • Updated all test cases to use the new function name
  • Added test for mixed content types (text + non-text) returning undefined
  • Changed test expectation for multiple text items from undefined to joined text

Implementation Details

The refactored extractAllTextContent() uses early returns for validation, then iterates through all content items to collect text values. If any item is not a text type, the function returns undefined immediately, ensuring the caller only uses this path when all content is text.

https://claude.ai/code/session_011WvveGQ76Mq6vapAWxamVt