Description

With validate_input_modes=True, a card that declares a media range such as image/* rejected every concrete image type, because the validator compared type/subtype literally. This was the question left open in #1425, and it breaks cards built by other SDKs: Microsoft Agent Framework's A2A hosting adapter documents image/* modes and returns a native AgentCard for use with this SDK's handlers, so the same card accepts image/png there and rejects it here (reproduction in #1468).

A declared type/* now matches any part of that type, and */* matches any part (RFC 9110 section 12.5.1). Everything from #1425 stays: case-insensitive, parameters ignored, parts without media_type unchecked. Exact declarations behave as before, and image/* still rejects video/mp4.

$ uv run pytest tests/utils/test_input_mode_validator.py   # fix reverted, new tests kept
5 failed, 18 passed
$ uv run pytest tests/utils/test_input_mode_validator.py   # with the fix
23 passed
$ uv run pytest --cov=src
src/a2a/utils/input_mode_validator.py   100%
TOTAL   94%
$ ./scripts/lint.sh
SUCCESS: All linting and formatting tasks complete!

The full local run was on Windows, where 4 failures and 8 errors in tests/integration/ (cross-version client/server, push notifications, samples smoke) also happen on unmodified main; they start subprocess servers and are unrelated to this change. CI covers them on Linux.

AI assistance: the analysis, the change and the tests were prepared with an AI coding assistant; I reviewed the diff and ran the checks above.

• Follow the CONTRIBUTING Guide.
• Make your Pull Request title in the https://www.conventionalcommits.org/ specification.
• Ensure the tests and linter pass (Run bash scripts/format.sh from the repository root to format)
• Appropriate docs were updated (if necessary)
  • The validate_input_modes docstring now states how media ranges match.

Fixes #1468 🦕