The T.R.U.S.T. check: deciding which jobs you can safely hand to your AI bots
bfrench · x · 2026-11-15
Sharing Michael Fenech's T.R.U.S.T. check — a plain-language framework for deciding how much supervision each job delegated to a Grok Bot team needs: which tasks still require human review, which can run with less oversight, and what to improve first.
The core is a prompt you paste into your Chief of Staff bot: it picks up to three recurring jobs you want to supervise less, reviews recent examples, actual outputs and corrections, and makes evidence-based calls. "Trust should be earned one job at a time" — you might let a bot organize your inbox while still approving every reply.
The author says they've implemented it on their own bot fleet (see companion article).
More from coding & agent
- Bot audit experiment: triage plus deep-read cuts input 64% while staying stable — bfrench · 2026-11-15
- AMA: A professional agent red-teamer on how AI agents fail in production — vivvvekredddit1122 · 2026-11-15
- Anthropic designer Nate Parrott on Claude Design workflows and 5 tips to avoid AI slop — petergyang · 2026-11-15
- Perplexity agent paints a macaw in 20,068 brushstrokes with a JS Canvas engine — jeffgrimes9 · 2026-11-15
- The 6 attack surfaces to test in every customer-facing agent — vivvvekredddit1122 · 2026-11-15
- Agent tests stayed green while it refunded a customer twice — tom_mathews · 2026-11-15