Org-wide AI use standard
You're the staff engineer on a 40-person platform org where every team has independently adopted AI coding agents (Cursor, Claude Code, Copilot) with no shared rules. Velocity is up but you're seeing inconsistent error handling, three different retry libraries appearing in one quarter, and a near-miss where an agent committed a hard-coded service token. Leadership asks you to write the org's first 'AI use standard.' What do you actually put in it, what do you deliberately leave out, and how do you make it stick without becoming the bottleneck everyone routes around?
evaluate_merge_gate(finding_lines: list[str], approved_dependencies: list[str]) → list[str][["secret=svc_token_abc","style=naming"],[]]out["secret_scan"][["new_dependency=retry-go","new_dependency=zap"],["zap"]]out["dependency_allowlist"][["style=indent","tool_choice=cursor","prompt_style=terse"],["zap"]]out[]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.