Cloud Agent - Preserve Structured Failure Context

[eshurakov]

Summary

• Preserve structured Cloud Agent failure codes, workspace subtypes, retryability, and safe messages across wrapper, queue, callback, result, and reporting boundaries.
• Correct provider, shutdown, workspace bootstrap, and retry-exhaustion classification while preventing raw process output or credentials from crossing service boundaries.
• Update callback consumers to prefer validated structured failures with legacy fallback during rolling deployments.

Verification

N/A

Visual Changes

N/A

Reviewer Notes

The shared failure contract is strict for producers and tolerant for callback consumers so mixed-version deployments retain legacy errorMessage compatibility.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

Incremental commit adds a single explanatory comment above WRAPPER_NO_OUTPUT_TIMEOUT_MS directly addressing the previously flagged inline question about the 330 000 ms timeout — no logic changes, no new issues.

Files Reviewed (1 file — incremental since 3ed4b7f)

• services/cloud-agent-next/src/session/agent-runtime.ts — comment added at line 41 explaining the 5-min provider timeout + 30 s grace rationale behind WRAPPER_NO_OUTPUT_TIMEOUT_MS = 330_000

---

Previous reviews: incremental (8 files, commit 3ed4b7f): No issues found. Incremental (4 files, commit 85a6e0f): No issues found. Full review (68 files, commit fdc51d5): No issues found.

---

_{Reviewed by claude-4.6-sonnet-20260217 · 198,835 tokens}

_{Review guidance: REVIEW.md from base branch main}

[RSO]

This was a chonker to review. Seems like a lot of mangling of error types, but overall the right side of the diff makes more sense than the left side :D