fix(spec): hold the identifier check to declared names, not to metadata.extensions

[uny]

Description

validate_catalogs_identifiers in specification/v1_0/test/run_tests.py walks every dict and list under a component or function definition and checks the keys of any properties object it meets against UAX #31 (str.isidentifier()). That reaches into metadata.extensions, which common_types.json's Extensions types as opaque vendor JSON — keys are identifiers, values are unconstrained (patternProperties: {…: {}}). The naming rule itself (docs/a2ui_protocol.md line 529) covers "component names, function names, and argument/property names": the names a catalog declares.

So a catalog component carrying

"metadata": {
  "extensions": {
    "com_example_legacy": { "properties": { "x-legacy-name": { "type": "string" } } }
  }
}

passes the structural (ajv) check against catalog_definition.json and is refused by the identifier check with Invalid argument/property name: 'x-legacy-name', although no stated rule constrains what is inside an extension value.

This PR:

1. factors the walk into iter_subschemas, which yields the direct subschemas of a schema through JSON Schema subschema positions only (properties, items, prefixItems, allOf/anyOf/oneOf, not, if/then/else, $defs, patternProperties, additionalProperties, dependentSchemas, contentSchema, … plus the legacy definitions/dependencies that ajv still evaluates under --spec=draft2020) and skips annotation values such as metadata, description, examples; validate_catalogs_identifiers checks properties keys through it;
2. adds exactly such an extension to test/testing_catalog.json as the fixture;
3. adds validate_identifier_walk, a harness step that places an invalid name under every keyword the walk must follow (and inside metadata.extensions / examples, where it must not look) and checks the exact names reported — the shipped catalogs only ever carry valid names, so without this a keyword dropped from the tables would go unnoticed.

Fixes #2705.

Reproduction

No workflow runs run_tests.py, so:

cd specification/v1_0/test && yarn install && python3 run_tests.py

• main + the fixture alone: structural check passes, identifier check fails with the message above (166 passed / 1 failed).
• This branch: 193 passed (167 + 26 synthetic cases).
• Non-regression is part of the run: a hyphenated key under allOf → items → properties, under a function's args.properties, under a declared property named metadata, and under each of the 20 subschema keywords is still refused, and dropping a keyword from the tables fails its case.

Pre-launch Checklist

One time:

• I signed the CLA.
• I read the Contributors Guide.
• I read the Style Guide.
• I have built and run at least one sample. — N/A, test-harness-only change

For this PR:

• I have updated the relevant CHANGELOG.md file. — none under specification/
• I updated/added relevant documentation. — none needed; behaviour now matches the prose
• My code changes (if any) have tests. — the fixture in testing_catalog.json and the synthetic cases in validate_identifier_walk
• If my branch is on a fork, I have verified that scripts/e2e_test.sh passes. — N/A, specification/v1_0/test only; python3 run_tests.py, pyink --check and prettier --check pass

🤖 Generated with Claude Code

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[gemini-code-assist]

Code Review

This pull request introduces a utility to recursively validate property names against UAX #31 identifiers within JSON schemas, including comprehensive synthetic test cases to ensure the identifier walk correctly traverses subschemas. The reviewer suggested an optimization for the iter_subschemas function to improve performance and readability by using a single dictionary mapping for keyword categories instead of multiple tuples. No review comments were removed.

[Varun-S10]

Hi @uny, I have reviewed and verified the changes. Thank you for your contribution.

Hi @gspencergoog, I would like to escalate this PR. Could you please take a look?