Skip to content

test(browser): record controlled Agent Task resource evidence - #72

Draft
seonghobae wants to merge 2 commits into
test/agent-task-semantic-role-name-evidencefrom
test/agent-task-resource-evidence
Draft

test(browser): record controlled Agent Task resource evidence#72
seonghobae wants to merge 2 commits into
test/agent-task-semantic-role-name-evidencefrom
test/agent-task-resource-evidence

Conversation

@seonghobae

@seonghobae seonghobae commented Aug 11, 2026

Copy link
Copy Markdown
Contributor

Partial implementation of #28, stacked on PR #71.

Buyer-visible gap

Issue #28 requires real browser/task RSS, observation bytes, action latency and task duration. PR #70/#71 prove real pinned-Chromium task execution plus browser-computed role/name evidence, but the runtime evidence previously lacked measured resource fields.

Dependency

This Draft is stacked on unchanged exact PR #71 head 5f1f972f3e9888faa44af184fd54a466d20b6ddb and must remain Draft while #71/#70/#65 are active.

TDD / RCA sequence

  1. Exact test-only head a9402a13c9ed429b8f3be2c623b994a0dfda3bb4 required strict Linux /proc/<pid>/status RSS parsing/sampling, ChromeDriver-issued goog:processID, and bounded browser_process_rss_bytes, semantic_observation_bytes, action_latency_ms, and task_duration_ms evidence.
  2. CI run 31454745237, Rust contracts job 93666110420, checked out that exact test-only head and failed at the intended repository contract because _parse_linux_proc_status_rss_bytes did not exist. This is the observed RED boundary.
  3. Current exact head 1a7186085abe926c1d0e5b22c36760965d6e237b implements only the bounded evidence boundary: strict positive single-VmRSS parsing with unit/shape/u64 overflow rejection; bounded /proc/<pid>/status reads for the ChromeDriver-issued browser PID; exact UTF-8 JSON byte size for the two controlled role/name observations; monotonic clear/type/click action latency; and end-to-end task duration. Invalid/missing process identity or malformed resource evidence fails closed.

Exact-head proof

On exact head 1a7186085abe926c1d0e5b22c36760965d6e237b against unchanged exact prerequisite #71 head 5f1f972f3e9888faa44af184fd54a466d20b6ddb:

  • CI run 31454903615: 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 31454903620, job 93666566904: success on Ubuntu 24.04 using Chrome for Testing 150.0.7871.129 / revision r1639810;
  • Agent Task repeats 3/3, pass rate 1.0;
  • measured browser-process RSS per trial: 215326720, 214568960, 213716992 bytes;
  • bounded semantic observation size: 95 bytes in each trial;
  • clear/type/click action latency: 133.282, 130.186, 110.554 ms;
  • end-to-end task duration: 1144.545, 957.991, 906.982 ms;
  • every trial also preserves semantic verification, exact post-condition, input echo, extensions-disabled execution and profile cleanup;
  • bounded evidence artifact mv3-compatibility-1a7186085abe926c1d0e5b22c36760965d6e237b, artifact ID 9087662526, uploaded artifact SHA-256 a7f8ec5ae716ed723e9dd7ec84eeac3478c30fddd6eb7c9bf0d411f1a8990ee5;
  • 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 lane is execution evidence only. browser_process_rss_bytes is intentionally narrower than task RSS: a single ChromeDriver-reported browser PID does not attribute Chromium renderer/GPU/utility descendants and must not be presented as whole-browser or whole-task RSS. Full multi-process attribution still belongs to a trusted adapter/process-set boundary that composes the existing #51/#66 resource contracts. semantic_observation_bytes measures only the bounded controlled role/name evidence serialized by this runner; it is not raw DOM size, model-context cost, source-node/network provenance, or product SemanticNodeObservation composition. No GPU/VRAM/local-AI evidence, production browser adapter, policy dispatch, or real-site compatibility is claimed.

@coderabbitai

coderabbitai Bot commented Aug 11, 2026

Copy link
Copy Markdown

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: bbfe4149-9964-4ae5-be08-e95ba97fb7e4

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

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

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant