aios-long-running-harness

# AIOS Long-Running Harness ## Overview Use this harness to keep long tasks stable under UI drift, model variability, and partial failures. It maps Anthropic's long-running-agent harness ideas into this repository's file-based workflow. ## Harness Loop 1. Preflight: lock objective, stop conditions, budgets, and required artifacts. 2. Plan: split into idempotent steps with explicit success/failure evidence. 3. Execute: run one step at a time with tool output capture. 4. Verify: assert completion from page evidence, not assumptions. 5. Checkpoint: persist current state, artifacts, and next action. 6. Recover: on failure, classify and retry only with a changed hypothesis. 7. Complete: run final verification and write summary doc. ## rex Command Boundary - The harness is an execution and recovery host, not a second software-workflow router. When a rex Workflow Activation exists, execute only the Provider selected by the current rex Command. - Return the Command's required Evidence to rex before advancing. Do not choose the next Capability in the harness or preload a fixed Provider chain. - Bundled `rex-*` Providers are the only software-workflow Providers. Do not enable compatibility substitutions or run an external playbook chain. - Dispatch parallel agents only after the AIOS policy selects one planned work item with independent domains; keep coupled or shared-state changes sequential. - Before claiming run success, apply the AIOS host verification gate and record concrete ...