autonomous-agent-loop

What problem does it solve?

This Skill prevents long autonomous coding sessions from going off-scope or ending without trustworthy evidence by keeping a goal-bounded controller loop with receipts, state persistence, and explicit stop conditions.

Core Features & Use Cases

• Goal-bounded loop control: Converts approved scope into a ready task graph, dispatches bounded workers, and continues until completion, blockage, pause, or stop.
• Evidence-first reconciliation: Records durable runtime receipts and workflow state, then reconciles outcomes and blocks completion unless required verification and reviewer coverage are present.
• Scope safety and drift guards: Enforces include/defer/reject/spike classification, compares plan/PRD hashes against a goal source drift guard, and prevents non-dry work without policy/readiness gates and user intent.

Quick Start

Use autonomous-agent-loop when you need the AI to keep executing a user-approved, goal-until-complete plan while producing evidence and safe stop behavior for a long worktree or multi-agent session.