Forward-Future/loop-library — the monorepo behind Loopy (duplicate of Forward-Future/loopy)

TL;DR

This capture is the Forward-Future/loop-library repository (2.2k★, MIT, JavaScript) — the monorepo that holds the Loop Library website/catalog (loop-library/: a site/ shell + a Cloudflare worker/ with a SQLite Durable Object) and the Loopy skill (skills/loopy/, with a loop-library compat alias). It is the same project already distilled as Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops, viewed from the repo-name angle instead of the skill-name angle: the same two parts, the same load-bearing "loop" idea (a one-shot prompt turned into a bounded, verifiable, repeatable workflow answering goal / verify-progress / use-the-learning / stop-or-ask), the same eight-path Loopy lifecycle (Discover, Find, Loop Doctor, Adapt, Craft, Run, Debrief, Publish), the same ≥2-occurrence rule, the same self-describing loop anatomy (Use when / Prompt / Verify / Steps / Notes / Related), and the same web plumbing (Cloudflare Worker, append-only revisions, fail-closed OAuth voting) that Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops already held as out-of-scope engine detail. Nothing is added over the canonical distillate — recorded here as duplicates (suppressed-but-logged, never silently dropped), with no new concepts and no concept edits. Read Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops for the substance; this node exists only so the second capture of the same project is accounted for.

What it is & how it is organized

• A GitHub repo, read at triage altitude. Two separable parts — identical to what Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops already maps:
  • loop-library/ — the website/catalog: site/ (static shell) + worker/ (Cloudflare Worker holding loop records in a SQLite-backed Durable Object, rendering JSON / Markdown / plain-text / llms.txt catalogs from one database write). The catalog is the library; no install needed to browse, and agents can read it directly.
  • skills/loopy/ — the installable skill (npx skills add Forward-Future/loopy … --agent {codex,cursor,claude-code}) that guides an agent through the loop lifecycle against the live catalog.
• The maintainer-notes appendix (database-backed publish/import/export/restore, fail-closed staged OAuth-voting rollout, HMAC nonce-bound state) is the web-app plumbing the prior distillate already logged under Held, not dropped — reconfirmed here, not re-distilled.

Concepts introduced

None. Every concept this capture embodies already has a page:

• Agent Loop — the bounded, feedback-driven workflow (four-question anatomy, real acceptance check, explicit stop conditions, hand-back to a human).
• Reusable Workflow Library — agent workflows as durable, cataloged, publishable, machine-consumable artifacts.

Held, not dropped

Same held themes as Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops — nothing new to add, listed so they are not silently dropped on this second capture:

• Cross-agent skill packaging / distribution — npx skills add portable across Codex / Cursor / Claude Code.
• Machine-readable agent catalogs — llms.txt, catalog.json, agent guide as the interface an agent reads instead of a human UI (touches Agent-Computer Interface (ACI)).
• Catalog publishing infrastructure — Cloudflare Worker + SQLite Durable Object, append-only revision per publish, draft/archive states, enforced bootstrap digest.
• Fail-closed staged rollout & OAuth voting security — HMAC nonce-bound state, no-store callback bridge, VOTING_UI_ENABLED flag flip after a smoke test.

Key claims

No new claims — all are already recorded on Agent Loop and Reusable Workflow Library via Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops. The load-bearing ones, unchanged:

• A loop is a playbook with feedback built in — repeatable, not one-shot. principle → Agent Loop.
• Loops are deliberately bounded — a real check, a finite run boundary before acting, and a hand-back to a human. principle → Agent Loop.
• Agent workflows are reusable artifacts worth naming, cataloging, and sharing. principle → Reusable Workflow Library.

Why this duplicates the existing graph

duplicates is the honest stance. The prior distillate Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops was already built from this exact Forward Future project (the Loopy skill is skills/loopy/ inside this repo; the loop-library skill name is an explicit compat alias). Its "What it is" already describes both parts, its concepts already exist, and its held themes already cover this capture's maintainer-notes plumbing. This capture surfaces no new concept, claim, or detail — so per firehose's constructive-dedup rule it is written and logged (never dropped) but marked duplicates: no concept created, no concept updated, no backlink churn. The canonical node is Forward-Future/loopy — a catalog and skill for bounded, reusable agent loops; consult it, not this one, for the substance. Had this been an independent source landing on the same claims it would be corroborates ("N sources agree"); it is not independent — it is the same repository captured twice.

Linked from

• This Week