Repository navigation
Add independently graded browser-agent repair pilot - #80
Merged
Merged
Conversation
|
View your CI Pipeline Execution ↗ for commit ce29c29
💡 Verify your cache is correct by running tasks in a sandbox. Read docs ↗ ☁️ Nx Cloud last updated this comment at |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
An agent returning a plausible patch is insufficient evidence that it fixed the intended UI behavior. This pilot gives full browser/source tools to an agent, captures its patch against an immutable seeded app commit, then applies that exported artifact to a fresh worktree/install for independent browser grading. A second condition also supplies a persisted Domscribe annotation and exact runtime/source tools.
The controlled cart defect updates both variants sharing an SKU. Acceptance checks only the intended line changes, reload persistence, sorting/filtering, Cancel, application errors, and functional production UI without instrumentation. Local execution builds Domscribe artifacts from a separate clean worktree at pinned main
2306385, records/validates their digest, and preserves the primary checkout. SDK execution adds only a selected Codex kit, existing provider-secret resource reference, and safe required extra artifacts to the existing bounded lifecycle runner. No credential values are read/copied, and no GitHub issue ingestion, publishing or deployment is implemented.Validation:
line-aand handled ambiguous source lookup via an exact instance.Cloud qualification is still blocked: both existing provider-secret references returned HTTP 409 failedPrecondition with no details, even for a shell kit. The native Codex kit starts without a binding but a minimal model request returns HTTP 401. All accepted diagnostic sandboxes were confirmed deleted. The user has been asked about provider/account setup. Keep this PR draft until the native cloud flow is qualified; local results do not establish cloud success.
One seeded task and one final local trial per condition do not establish demand or efficiency. Completed MCP calls were 5 baseline / 15 enriched; the enriched prompt requires additional Domscribe work and strategies differ. Exact model ID was not reported. This is feasibility evidence with an explicit cloud gate, not a performance claim. Usage, evidence and limits are in
docs/cloud-agent-repair-pilot.md.