PPLSBROOpenAI Build Week · H1

Trusted agent execution

See every boundary.
Trust every result.

Approve a visible task contract, run a platform-owned agent in one deny-all sandbox, and inspect the verification, usage, and audit evidence behind the result.

Execution boundary
BrowserapprovalControl planefilesSandbox
Network: deny-all1 sandbox≤ 1 model call
Hackathon demonstrationPlatform-owned runner · synthetic data · not production sandbox certification
01

Review the contract

csv_cleanup_demo_v1

Normalize and deduplicate a synthetic customer CSV

Synthetic input
Cust. Co.ContactStatus
Acme LabsAda Lovelace active
Northstar LLCGrace HopperACTIVE
Acme LabsAda LovelaceActive

Permissions

  • Read the fixed synthetic input
  • Write cleaned.csv and reconciliation.json
  • Request one column_mapping_v1 decision
  • No network access and no secrets

Success criteria

  • Required output columns are present
  • Status values are normalized
  • Duplicate rows are removed
  • Input and output row counts reconcile
  • Only declared agent artifacts are accepted
Agent artifactscleaned.csv · reconciliation.json
Platform evidenceaudit-receipt.json · verification result · model usage and cost