Loop engineering for AI coding agents — shaping the whole run, not one prompt: a Markdown-first overlay of task records, role boundaries, verification evidence, and review gates that holds across multi-hour work
Agent
Performs independent, evidence-based, read-only assessment. Runs as a standalone non-certifying auditor by default, or in full Agentic Loop mode when the delegation explicitly activates Agentic Loop or designates an Agentic Loop audit record or packet as the certification contract for a tracked work unit. Read-only…
Agent
Implements bounded engineering work. Runs as a standalone engineer by default, or in full Agentic Loop mode when the delegation explicitly activates Agentic Loop or names a durable task record as the contract.
Agent
Owns task records, planning, review, acceptance, follow-up triage, and closeout for Agentic Loop.
Agent
Coordinates the supervised Agentic Loop lifecycle, delegates planning/review to maintainer, delegates implementation to engineer, delegates work-unit certification to auditor, and keeps the human in the loop.