Source: raw/x-bookmark-2067682870009925736.md + raw/x-bookmark-2067742219235471419.md (@MatthewBerman, 2026-06-18) — launch posts for the Loop Library at signals.forwardfuture.ai/loop-library/. Bodies extracted verbatim from the bookmarks; the catalog’s internal contents are not independently verified here.

A curated, community-submittable catalog of agent loops — reusable “prompt the agent, read the output, reprompt until done” patterns — launched by Matthew Berman under the Forward Future brand. It is a directory companion to the wiki’s existing agent-loops reference material: where Loop Engineering catalogs the concept space, the Loop Library is a browsable, contribute-your-own index of concrete loops.

Key Takeaways

  • What it is. “A curated list of agent loops you can use right now. Find loops, submit your own.” Hosted at signals.forwardfuture.ai/loop-library/; launched 2026-06-18 (1,327 likes at capture).
  • It is a directory, not a framework. The value is the catalog of named, copy-pasteable loop prompts plus a submission path for the community — the same “loops as a unit of work” framing in The Loop Is the Unit of Work.
  • Worked example — the “architecture-satisfaction loop” (credited to Peter Steinberger, @steipete): “Refactor until you are happy with the architecture. After each significant step, live-test the system, run autoreview, and commit. Track progress in /tmp/refactor-{projectname}.md.” It pairs an open-ended satisfaction condition with a per-step verify-and-commit cadence and an external progress file — a textbook verifier-first loop.
  • Why it matters for this wiki. It is a live, growing source of vetted loop patterns to mine for the agent-loops topic — and a place to contribute the loops documented here.

Try It

  1. Browse signals.forwardfuture.ai/loop-library/ for a loop matching your task (refactor, research, migration, review).
  2. Try the architecture-satisfaction loop on a messy module: instruct the agent to refactor until the architecture satisfies a stated bar, live-testing + autoreviewing + committing after each step, and tracking progress in /tmp/refactor-<project>.md.
  3. Submit loops that work for you back to the Library so the catalog compounds.