Loop Library

Loop Library is a free field guide from Forward Future that gathers repeatable AI agent prompts — called loops — for engineering, research, evaluation, and operations work. Each loop pairs a plain instruction with a verification step and an explicit stopping condition, so an agent knows when the task is actually finished rather than running indefinitely. The collection is browsable, searchable, and open to community submissions.

What It Is

A curated, copy-and-paste library of agent prompts. The homepage states that a useful loop specifies five things: a trigger, an action, proof, memory, and a stopping condition. Every entry in the table includes the prompt itself plus a paired verify-and-stop note describing the evidence that proves the work is done.

How Loops Are Structured

• Action — The instruction you hand to an agent, written to be copied directly.
• Verify / stop — The condition that confirms success and tells the agent to halt, such as a passing test suite or a clean log.
• Attribution — The practitioner credited with the original workflow.

What's In the Library

The current set leans toward agentic engineering: an overnight documentation sweep, a full test-coverage loop, a sub-50 ms page-load loop, a production error sweep, an SEO/GEO visibility audit, repository cleanup, and batch release loops, among others. The list is dated and refreshed periodically, with new loops added over time.

Who Contributes

Loops are attributed to named practitioners, including Matthew Berman and Peter Steinberger. Forward Future says every submission is reviewed before publishing, and contributors are asked to confirm they have the right to share and to leave out secrets or private code.

Weekly Picks

The site offers an optional weekly email that highlights one loop at a time, framed as one useful loop, once a week.