Review Output as Future Maintainer
AGENT-REVIEW-OUTPUT-AS-FUTURE-MAINTAINER
Summary
Review agent output as someone who will not have the chat transcript. This keeps attention on durable code, docs, tests, validation proof, and residual risk.
Rule
Review agent output from the perspective of a future maintainer who did not see the session.
Why
Agent output can pass checks while depending on prompt-only context, vague names, hidden state, over-broad abstractions, missing tests, or undocumented behavior changes. The future maintainer will not have the chat transcript in front of them. They need the code, docs, tests, and handoff artifact to explain the change without replaying the session.
Reviewing as a future maintainer keeps attention on the durable artifact: whether the owning module is clear, behavior is observable, edge cases are covered, documentation is truthful, and residual risk is named.
Helps
- Catches agent output that is locally plausible but hard to maintain after context disappears.
- Turns review from confidence assessment into durable artifact inspection.
Limits
Do not require perfection before accepting a useful, coherent chunk. If the artifact is correct but has follow-up cleanup, name the residual risk and keep the follow-up scoped. Apply deeper review to public APIs, security-sensitive work, generated behavior, persistence, rendering, and changes that will be hard to reverse.
Agent Instruction
Review agent output as a future maintainer: check correctness, edge cases, API clarity,
documentation truthfulness, readable ownership, focused tests, validation proof, and residual risk.Mechanisms
Supported by review packets, focused validation logs, owning-module inspection, code-review checklists, generated-drift checks, screenshots for rendered output, and handoff notes that separate proof from residual risk.