ScaffoldBench

RunSonnet 5.5repeat 1

Orchestrator-workers on Triage a new service desk ticket

claude-sonnet-5-5__orchestrator-workers__support-easy__r1

Score

1.000

passed
design ran as designed
Cost
$0.050
Latency
21.2s
setup 0.0s, grading 0.0s
Turns
11
Tool calls
7
Subagents
1
Tokens in / out
28k / 2.1k
19,350 read from cache

Did the design run?

every check held
  • Run completed: heldharness.run-completedThe environment was built, the architecture returned and grading finished, none of them throwing.
  • No API failures: heldharness.apiNo model call failed at the API.
  • Every call ended: heldharness.calls-endedEvery call ended with a result, or was stopped by the run's own limits.
  • No harness tool denied: heldharness.harness-toolsThe sandbox guard never denied a harness tool, nor a tool the session itself offered.
  • Tool servers connected: heldharness.mcp-connectedEvery tool server the harness gave a call was connected when the session started.
  • No foreign tool servers: heldharness.mcp-hermeticNo call loaded a tool server the harness did not give it.
  • Builtin tools: heldharness.builtin-toolsA check recorded with the run.
  • Structured replies arrived: heldharness.structured-outputEvery call that asked for structured output and ran to the end got it.
  • No silent fallbacks: heldharness.not-degradedNo phase fell back on a default (an empty plan, a default winner) instead of its output.
  • The task was decomposed: heldorchestrator-workers.decompositionThe orchestrator’s reply was read through its schema and held at least one subtask.
  • The fan-out happened: heldorchestrator-workers.fan-outEvery subtask got a worker, every worker ran, and the recorded subagent count matches.
  • The synthesis wrote the answer: heldorchestrator-workers.synthesisThe final answer came from the synthesis call, not from a worker’s report standing in for it.
  • Each call was shown the right tools: heldorchestrator-workers.toolsThe orchestrator and the synthesis were shown no tools and every worker the task’s full toolset.

Phases

Decompose: 0.0s to 5.0s, finished, 0 tool calls, $0.011
Worker 1: all: 5.0s to 14.3s, finished, 7 tool calls, $0.024
Synthesize: 14.3s to 21.2s, finished, 0 tool calls, $0.015

Drawn from the phase records: timing and cost per call, no tool calls or messages.

PhaseDurationCostTurnsToolsModel
Decompose5.0s$0.01120claude-sonnet-5-5
Worker 1: all9.3s$0.02487claude-sonnet-5-5
Synthesize6.9s$0.01510claude-sonnet-5-5

The call-by-call replay is not published for this run. Traces are large, so the repository keeps 9 of this run set’s 900 and leaves the rest out; the score, phases and final answer on this page come from the committed run record.

Runs you can replay

Record this cell yourself: npm run bench -- run --strategies orchestrator-workers --tasks support-easy --models claude-sonnet-5-5 --repeats 1

Score breakdown

  • passSeverity matches the escalation matrix for the service and impact scope
    1.00weight 0.27

    severity=sev3 (matrix says sev3)

  • passSLA deadline is the triage time plus the response target for that severity
    1.00weight 0.27

    sla_due_at=2026-03-17T17:20:00Z (expected 2026-03-17T17:20:00Z)

  • passAssigned to the engineer on call for the owning rotation at the deadline
    1.00weight 0.27

    assignee=E-301 (expected E-301)

  • passTicket left in status open
    1.00weight 0.09
  • passEscalation matrix read before the severity was set
    1.00weight 0.09

Penalties

No penalties were applied.

Scorer metadata

tool calls
7
failed tool calls
0
distinct tools
7

Final answer

I triaged T-1041 as described below, but this comes from a single worker report that I couldn't check against the ticket system myself.

  • Severity: The worker set it to sev3. The escalation matrix gives VPN with team-wide impact as sev3, and the ticket's scope is "team" (the Berlin sales team, about nine people).
  • Deadline: The ticket had none, so the worker set 2026-03-17T17:20:00Z. That is a triage time of 09:20Z plus the sev3 response target of 480 minutes.
  • Assignment: The worker assigned it to E-301 Yuki Sato. She is on call for network-primary, the rotation that owns VPN, and her 08:00–20:00Z shift covers the 17:20Z deadline. The ticket is now open.
  • Escalation: None. The matrix names no escalation rotation for sev3, and the task didn't ask for one.
  • Other tickets: The worker reports leaving every other ticket unchanged.

Date mismatch: The deadline date (2026-03-17) doesn't match today's date in my environment (2026-10-06). The ticket system's clock may simply differ from mine, but if the 09:20Z triage time was taken from the wrong source, the deadline and Yuki Sato's on-call coverage would need rechecking.

Severity assumption: The severity depends on the impact scope being "team". If the outage is wider, it should be re-graded against the matrix.

Provenance of the numbers on this page

run set
scaffold-or-scale-v1
models
Haiku 4.5, Sonnet 5.5, Opus 5.5
effort
high
repeats
2
fixtures
replay
runs
900 / 900
strategy cost
$354.44
judge cost
$9.19
wall clock
9h 48m
git
fce5807b2d

Cost excludes judge calls, which are tracked separately so a scoring model never flatters or penalises an architecture. Every figure on this site is read from results/ at build time.