test(browser): verify semantic role and name before controlled action

[seonghobae]

Partial implementation of #28, stacked on PR #70.

Buyer-visible gap

PR #70 proves real pinned-Chromium input/click/post-condition execution for the controlled Agent Task, but its controlled targets are located only by fixture CSS and the branch explicitly does not claim semantic role/name evidence. This lane adds a bounded browser-computed accessibility evidence check before either controlled action proceeds.

Dependency

This Draft is stacked on unchanged exact PR #70 head f9917cdd8050c9fdf0aefa669f4d981af85479d6 and must remain Draft while #70/#65 are active.

TDD / RCA sequence

1. Exact test-only head 977d2682dc191ca6b26b9de631a3642680abdbc0 required _get_element_semantics, W3C WebDriver computed-role/computed-label endpoints, exact textbox / Task text and button / Submit task semantics, and credential-free evidence booleans.
2. CI run 31454219111, Rust contracts job 93664601520, checked out that exact test-only head and failed at the intended repository contract because _get_element_semantics did not exist. This is the observed RED boundary.
3. Current exact head 5f1f972f3e9888faa44af184fd54a466d20b6ddb adds only the bounded semantic read/verify step around the existing controlled elements. The runner still uses CSS solely to obtain the synthetic fixture element handle, then reads browser-computed role/name through WebDriver before clear/type or click. Mismatched or malformed semantics fail closed with bounded credential-free errors.

Exact-head proof

On exact head 5f1f972f3e9888faa44af184fd54a466d20b6ddb against unchanged exact prerequisite #70 head f9917cdd8050c9fdf0aefa669f4d981af85479d6:

• CI run 31454448709: success, including Python repository contracts, canonical formatting, locked workspace/all-target check, full tests, strict Clippy, rustdoc and exact owned production function/line/region/branch coverage;
• pinned real-browser run 31454448710, job 93665257103: success using Chrome for Testing 150.0.7871.129 / revision r1639810;
• Agent Task repeats 3/3, pass rate 1.0; every trial reports input_semantics_verified=true, submit_semantics_verified=true, post_condition=true, input_echo_verified=true, extensions_disabled=true, and profile_cleaned=true;
• bounded evidence artifact mv3-compatibility-5f1f972f3e9888faa44af184fd54a466d20b6ddb, artifact ID 9087505678, upload SHA-256 81a0fe4190a413a088a9fd8b3094367966b84c77770c8a7fce1f85bba3c888e5;
• CodeRabbit exact-head commit status: success; and
• GitHub reports the Draft mergeable against its unchanged prerequisite base.

No predecessor-head result is transferred.

Truth boundary

This remains executable CI evidence, not the final OriginWeave browser runtime. CSS selectors remain controlled test-harness locators; verifying browser-computed role/name does not itself implement role/name search, a versioned WebDriver BiDi/CDP product adapter, OriginWeave SemanticNodeObservation composition, opaque node-handle registration, typed policy dispatch, source-node/network provenance, resource telemetry, or real-site compatibility. Those remain separate boundaries of issue #28.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d1339923-35ac-440f-a05a-2bd4720fbb63

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}