Add Copilot Studio SDK + Direct Line GPT 5.5 @ 256x leaderboard entries

[c-jatangithanuja]

Summary

Adds two Enterprise-Bench leaderboard entries for Copilot Studio + GPT 5.5 @ 256x — same agent/model/prompt, two Harbor backends:

File	Backend	pass@1	pass@10	Harbor job
2026-08-01__copilot-studio-sdk__gpt-5-5.yaml	SDK	17.1%	35.7%	https://hub.harborframework.com/jobs/ed1ec192-cf74-48ca-8840-006441240db6
2026-08-02__copilot-studio-direct-line__gpt-5-5.yaml	Direct Line	25.7%	50.0%	https://hub.harborframework.com/jobs/971f63dd-b0b0-46bf-902f-260edfc92bd5

Both runs: 14 tasks × k=10 = 140 trials, 0 errors. Jobs are public on Harbor Hub.

Type of contribution

• Agent result submission
• Task addition or update
• Dataset addition or update
• Documentation
• Bug fix / setup improvement
• Other

Validation

• make validate passes (or run before opening PR — see Commands)
• I ran at least one affected task, or explained why not below
• Dataset changes are synthetic and safe to publish (N/A — no dataset changes)
• New/changed task criteria avoid answer leakage (N/A — no task changes)
• Documentation is updated (N/A — leaderboard entries only)

Commands run

# Original benchmark runs (Aug 1–2, 2026)
# Model: GPT 5.5 — set in Copilot Studio portal (not Harbor CLI)
# MCP: unified gateway on :8010 via ngrok; see harbor/docs/copilot-studio-runbook.md

# Reproduce (SDK) — configs/256x.env: COPILOT_STUDIO_BACKEND=sdk, device_code auth
cd harbor && source .venv/bin/activate
set -a && source configs/256x.env && set +a
harbor run -p tasks/l1-l2-bench-v1 \
  -a agents.copilot_studio:CopilotStudioAgent \
  --env-file configs/256x.env --yes -k 10 -n 3 \
  --job-name "$(date +%Y-%m-%d__%H-%M-%S)__copilot-studio-sdk-gpt-5-5__256x" \
  --jobs-dir jobs/copilot-studio/256x/full-k10-sdk3

# Reproduce (Direct Line) — set COPILOT_STUDIO_BACKEND=direct_line in configs/256x.env
harbor run -p tasks/l1-l2-bench-v1 \
  -a agents.copilot_studio:CopilotStudioAgent \
  --env-file configs/256x.env --yes -k 10 -n 3 \
  --job-name "$(date +%Y-%m-%d__%H-%M-%S)__copilot-studio-direct-line-gpt-5-5__256x" \
  --jobs-dir jobs/copilot-studio/256x/full-k10-direct-line

# Validate enterprise-bench repo (before PR)
cd enterprise-bench && make validate

[nimit2801]

Thanks for the leaderboard submission. First-touch read: this is a result-only contribution adding two Copilot Studio + GPT 5.5 @ 256x entries, one SDK backend and one Direct Line backend.

Current checks: CI / validate is passing, and the PR is mergeable. Risk looks limited to leaderboard metadata/result integrity rather than task, dataset, or code behavior changes.

Maintainer verification points:

1. Confirm both Harbor source jobs are public and match the submitted backends: SDK job ed1ec192-cf74-48ca-8840-006441240db6 and Direct Line job 971f63dd-b0b0-46bf-902f-260edfc92bd5.
2. Verify the reported scores in the YAML entries match the public jobs: SDK pass@1 17.1%, pass@10 35.7%; Direct Line pass@1 25.7%, pass@10 50.0%.
3. Spot-check that agent/model/prompt/config metadata is consistent across the two entries and clearly distinguishes SDK vs Direct Line.