Skip to content

test(browser): fail closed on oversized Agent Task observation - #128

Draft
seonghobae wants to merge 4 commits into
test/agent-task-structured-value-evidencefrom
test/agent-task-observation-bound
Draft

test(browser): fail closed on oversized Agent Task observation#128
seonghobae wants to merge 4 commits into
test/agent-task-structured-value-evidencefrom
test/agent-task-observation-bound

Conversation

@seonghobae

Copy link
Copy Markdown
Contributor

Partial implementation of #28, stacked on unchanged exact PR #105 head bc1d22d6c4848a173c55fdd18054574299488067.

Buyer/security gap

The pinned-Chromium controlled Agent Task measures its semantic-observation JSON byte size but only rejects the impossible zero-length case. Issue #28 requires oversized observation/text to be truncated or fail closed. Without an explicit byte ceiling at the executable browser-evidence boundary, later fixture growth or adapter expansion could silently turn an already bounded role/name snapshot into unbounded evidence.

Dependency

This Draft targets exact #105 head bc1d22d6c4848a173c55fdd18054574299488067 and must remain Draft while #105/#100 and their prerequisites remain active. No prerequisite check, review, status, artifact, coverage, or mergeability result transfers.

TDD state

Exact test-only head 71e78784332b2cbe34d3d704b6a03d0d69ff35c2 defines the missing executable bound before runner support exists. The intended contract requires:

  • one explicit finite MAX_AGENT_TASK_SEMANTIC_OBSERVATION_BYTES ceiling;
  • canonical compact UTF-8 JSON measurement;
  • exact-limit acceptance and one-byte-over fail-closed rejection;
  • rejection of empty and non-object observation values; and
  • the real controlled Agent Task browser pass to use the bounded measurement helper rather than an unbounded direct len(json.dumps(...)) path.

Production-like runner support is intentionally absent at this head so CI can establish the realistic RED.

Truth boundary

This lane bounds only the controlled semantic-observation evidence emitted by the pinned-browser compatibility runner. It does not define the final product-wide semantic snapshot budget, truncate browser content, authenticate a browser adapter, authorize model context, expose raw HTML, add arbitrary JavaScript, or complete the end-to-end OriginWeave browser adapter. A later production adapter must apply its own reviewed structured-observation bounds at the real observation boundary.

@coderabbitai

coderabbitai Bot commented Aug 12, 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: 854d174d-a948-4083-a647-57b502888012

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