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. ## Progress self-report and replan cadence Stuckness is self-reported by you in structured observations, never judged by a second model call and never guessed from message keywords. After every Execute/Verify round, emit: - `progress_made: true|false` — `true` only when the round produced new evidence, a new artifact, or a state change versus the previous round. Repeating the same tool call with the same result is `false`. - `blocked_reason: <empty|one line>` — empty when progress was made; otherwise the concrete blocker (selector gone, auth lost, tool error, waiting on input). Rules: - The harness counts consecutive `progress_made: false` rounds (threshold 3 unless the run preflight set another). At threshold, stop repeating: run Recov...