(MOT-4277) fix(harness): treat recovered function errors as quality

[ytallo]

Refs MOT-4277

Summary

• move shell/coder/sandbox function-call errors from a hard gate into an explicit quality criterion
• keep worker setup, coder workflow, exact host output, sandbox lifecycle, operation volume, and cleanup blocking
• lower the scenario floor to 50 so a fully validated recovered execution passes while losing 45 quality points
• preserve the 77.5 aggregate for one clean execution and one recovered execution

Why

A recoverable function-call error should expose poor execution quality without overriding independently verified final effects. Unrecovered errors still fail through the hard gate for the affected outcome.

Validation

• cargo test --locked --manifest-path harness/Cargo.toml -p harness-e2e
• cargo fmt --manifest-path harness/Cargo.toml --all -- --check
• git diff --check

Summary by CodeRabbit

• Documentation

  • Clarified how recovered-function errors affect shell_coder_sandbox quality scores and reports.

• Tests

  • Updated scenario scoring to distinguish verified outcomes from execution quality.
  • Preserved credit for objectively validated results even when recoverable errors occur.
  • Added coverage for score thresholds, weighting, and execution-quality behavior.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
workers	Ready	Preview	Jul 30, 2026 1:18pm
workers-tech-spec	Ready	Preview	Jul 30, 2026 1:18pm

[github-actions]

skill-check — worker

0 verified, 49 skipped (no docs/).

Layer	Result
structure	✓
vale	✓
ai	✓
render	✓

Four for four. Nicely done.

[coderabbitai]

Warning

Review limit reached

@ytallo, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 58 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 8ab8ab68-c455-41de-a512-aab61dd3cfc8

📥 Commits

Reviewing files that changed from the base of the PR and between 70e8cdd and 0c1532e.

📒 Files selected for processing (2)

• harness/tests/e2e/README.md
• harness/tests/e2e/src/scenarios/shell_coder_sandbox.rs

📝 Walkthrough

Walkthrough

The shell_coder_sandbox scenario now treats recovered function-call errors as an execution-quality signal, while preserving awards for objectively validated outcomes. Scoring weights, threshold documentation, evaluation logic, and unit tests were updated.

Changes

Shell coder sandbox scoring

Layer / File(s)	Summary
Define execution-quality scoring harness/tests/e2e/src/scenarios/shell_coder_sandbox.rs, harness/tests/e2e/README.md	Adds the execution-quality criterion, rebalances scoring weights, sets the pass threshold, and documents recovered-error behavior.
Apply execution-quality awards harness/tests/e2e/src/scenarios/shell_coder_sandbox.rs	Preserves validated awards, awards execution quality only without function-call errors, and tests threshold, weight, and award behavior.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• iii-hq/workers#635: Also updates shell_coder_sandbox scoring, evaluation, and documentation.

Suggested reviewers: andersonleal

Poem

I’m a bunny with scores in my paws,
Errors now signal quality, not pause.
Validated work still earns its due,
Clean calls add points to the queue.
Hop, test, and report—
The sandbox keeps score!

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/harness-e2e-recoverable-errors

---

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.}