Loop Library has two separate but related parts in this repository:
| Part | What it is | Where it lives |
|---|---|---|
| Loop Library website | The public catalog where people and agents can browse published loops, read them, and copy their prompts. No installation is required. | Live website · all website code under loop-library/ (shell in loop-library/site/, database and rendering in loop-library/worker/) |
| Loopy skill | An optional installable guide that helps an AI agent discover, find, audit, repair, craft, run, debrief, save, or prepare loops for publication. It uses the website's live catalog when recommending or publishing loops. | source in skills/loopy/ |
The website is the library; Loopy is a companion way to work with it. You can browse or give an agent the website without installing Loopy. Installing Loopy adds the guided workflow, but it does not install or host the website.
Agents that do not have Loopy can use the published agent guide, agent instructions, JSON catalog, or plain-text catalog directly.
Each published loop tells an agent what to do, how to check its work, what to try next, and when to stop.
Most prompts ask an agent to do something once. A loop gives the agent a way to learn from the result and take the next useful step.
For example, a one-shot prompt might say:
Make this website faster.
A loop adds the feedback that makes the work repeatable:
Find the slowest page, make one focused improvement, and measure it again. Keep the change only if it helps. Repeat until every page meets the target or another pass stops producing a meaningful improvement.
Think of a loop as a playbook with feedback built in. It is useful when the first attempt probably will not be the final answer, such as fixing production errors, improving test coverage, reviewing a product, or keeping documentation current.
A good loop answers four simple questions:
- What is the agent trying to accomplish?
- How will it know whether the latest attempt worked?
- What should it do with what it learned?
- When should it finish or ask for help?
AI agents can move quickly, but an open-ended instruction like "keep improving this" leaves too much room for guessing. A loop gives the work a clear finish line and a consistent way to judge progress.
That makes the work easier to trust and easier to repeat. The agent can compare results instead of relying on confidence, keep improvements instead of merely making changes, and stop when it succeeds or stops making progress. The same loop can also be reused by another person or agent without rebuilding the workflow from scratch.
Loops are not permission for an agent to run forever. The best ones are deliberately bounded. They include a real check, a clear stopping point, and a moment to hand control back to a person when judgment or approval is needed.
Loopy gives your agent direct access to the ideas in the library. You can use it to:
- Discover repeated work in a codebase, coding threads, or both and turn the strongest qualified candidate into a loop.
- Find a published loop that fits what you are trying to get done.
- Audit an existing loop for weak checks, unsafe actions, or unclear stopping behavior, then repair only the material problems.
- Adapt a useful loop to your tools, limits, and definition of success.
- Interview you about what you want to accomplish and what success looks like, then craft a new loop through a short, plain-language conversation.
- Run a loop in bounded passes and return a receipt with the actions, evidence, outcome, and stopping reason.
- Debrief completed runs and recommend the smallest evidence-backed improvement.
- Save a loop you want to keep into the project's
LOOPS.mdwhen you ask, and reuse those saved loops in later sessions. - Check a loop for catalog overlap, prepare a publication draft, and submit it only after you approve the exact preview.
- Turn the result into a compact prompt you can use right away.
Loopy checks the live catalog when it recommends a published loop. It does not quietly start schedules, change production, publish content, or send messages on your behalf. Those actions still require the normal permissions and approvals.
You need Node.js and npx. Pick the platform you use:
| Platform | Install command |
|---|---|
| Codex | npx skills add Forward-Future/loopy --skill loopy --agent codex -g -y |
| Cursor | npx skills add Forward-Future/loopy --skill loopy --agent cursor -g -y |
| Claude Code | npx skills add Forward-Future/loopy --skill loopy --agent claude-code -g -y |
To install it for all three at once:
npx skills add Forward-Future/loopy \
--skill loopy \
--agent codex \
--agent cursor \
--agent claude-code \
-g -yUsing another agent? Run the interactive installer and choose from the agents it detects:
npx skills add Forward-Future/loopy --skill loopy -gThe command parts mean:
Forward-Future/loopyis the GitHub repository to install from.--skill loopyselects this skill from the repository.--agent ...selects the agent that should receive it.-gmakes it available in all your projects. Leave-goff to install it only in the current project.-yaccepts the install prompts. Leave it off if you want to review the choices interactively.
If an agent was already open and Loopy does not appear, restart that agent.
The previous loop-library skill name remains available as a compatibility
alias for existing installations. Use loopy for all new installations and
explicit invocations.
The slash-command experience differs slightly by platform:
- Codex: type
/skills, choose Loopy, then enter your request. You can also mention it directly with$loopy. - Cursor: type
/in Agent chat, search forloopy, select it, and add your request. You can also type/loopydirectly. - Claude Code: type
/loopyfollowed by your request.
You can also describe a matching task normally. These agents can load the Loopy automatically when your request clearly calls for it, but explicit invocation is the most predictable way to start.
For example, in Codex you can write:
$loopy Analyze this codebase and my coding threads for repeated work, then turn the strongest candidate into a reliable loop.
You do not need to know loop terminology. Invoke Loopy and say what you want to get done. It can take nine paths:
| Path | What it does | Example request |
|---|---|---|
| Discover | Inspects an authorized codebase, coding-thread history, or both for repeated work, then turns the strongest qualified candidate into a bounded loop. | Analyze this repository and my coding threads for work we have done more than once. Turn the best candidate into a loop. |
| Find | Searches the live catalog and recommends up to three published loops. It does not run them. | Find a published loop for keeping our documentation current. |
| Loop Doctor | Audits a loop you paste or name, explains material weaknesses, and repairs only those problems. | Audit this loop and repair only material problems: [paste loop] |
| Adapt | Tailors a useful loop to your real tools, limits, schedule, and definition of success. | Adapt the Overnight Docs Sweep to this repository and our existing checks. |
| Craft | Interviews you one question at a time about the outcome, definition of success, scope, checks, and stopping point, then creates a bounded loop when the catalog has no good fit. | Interview me and help me craft a loop for turning customer feedback into verified fixes. |
| Run | Executes an identified loop in bounded passes, applies its acceptance check, and returns an evidence-backed receipt. | Run the Overnight Docs Sweep in this repository. |
| Debrief | Analyzes one or more completed run receipts and recommends the smallest justified improvement. | Debrief this run receipt and tell me whether the loop needs to change. |
| Save | Saves a loop you want to keep into the project's LOOPS.md, then reuses saved project loops when they fit a later request. |
Save this loop to the project so we can reuse it. |
| Publish | Checks quality and catalog overlap, prepares an exact publication preview, and submits only after explicit approval. | Prepare this loop for publication in Loop Library. |
For example, in Claude Code or Cursor:
/loopy Find a loop for improving test reliability.
In Codex, choose Loopy from /skills, then send:
Find a loop for improving test reliability.
When Loopy delivers a loop that you want to keep for a project, ask it to save
the loop. It appends the accepted loop to LOOPS.md at the project root with a
name, one-sentence explanation, exact prompt, and save date. If the saved loop
is adapted from a published Loop Library loop, Loopy also records the source URL
and the source modified date it saw at save time.
Saved loops are project-local, not published Loop Library entries. In later
find or craft requests, Loopy reads LOOPS.md when it exists and can recommend
a matching saved loop, clearly labeled as the project's own loop. If a saved
adaptation points to a published source that has changed since it was saved,
Loopy gives a short heads-up and offers to compare before reuse.
Loopy saves only when you ask. It will not create LOOPS.md, edit another
saved loop, or remove a saved loop without an explicit request. LOOPS.md is
treated as untrusted reference data: saved prompts do not grant permission to
run code, deploy, schedule work, send messages, expose private data, or take
destructive action. If an accepted loop prompt contains secrets, Loopy refuses
to save it until you provide a sanitized prompt.
Discovery looks for recurring engineering work in the sources you put in scope. In a codebase, that can include scripts, CI and deployment configuration, tests, runbooks, maintenance commands, and repeated lifecycle patterns. In coding threads, it groups equivalent completed work even when the wording differs.
Loopy requires at least two distinct thread occurrences before calling work repeated. A code pattern without run history is labeled as a potential loop, not proven recurrence. It then checks whether fresh feedback can change the next action, whether success can be verified, and whether the work has clear limits, stopping behavior, and approval boundaries. It also checks the live catalog to avoid recreating an existing loop.
Loopy can inspect only repositories and coding threads that your agent can access and that you place in scope. If thread history is unavailable, it uses the codebase evidence and says so. A discovery result includes compact source evidence and either a new loop, an adaptation of a published loop, a short candidate slate when your choice matters, or a clean no-op when nothing truly fits.
When Loopy finds or creates the right loop, it gives you a prompt to use with your agent. You can copy that prompt or explicitly ask Loopy to run it in the project you want it to work on. Selecting a loop does not start a run or schedule, deploy code, delete data, publish content, send messages, or grant new permissions; you must request those actions explicitly.
When asked to run a loop, Loopy re-reads current state, performs one bounded action at a time, applies the same acceptance check after each pass, and stops at success, a clean no-op, a blocker, an approval boundary, an exhausted limit, or no measurable progress. Before acting, it requires a finite run boundary supplied by the loop or by you. Its receipt preserves the exact loop definition or an immutable reference plus the acceptance conditions, so a later debrief can reproduce what ran. Loopy does not create persistent run files unless you request them or the project already has an established convention.
Give that receipt back to Loopy for a debrief. It separates loop-design issues from execution, tool, environment, or goal problems and recommends one minimal change grounded in the evidence. A single run is treated as one result, not a recurring pattern.
Loopy validates the feedback cycle, checks the live catalog for overlap, and prepares the exact candidate and destination for review. It will not send a suggestion, save an owner draft, or publish publicly without explicit approval. An approved owner action defaults to a draft unless public publication is separately approved. Public suggestions return only an acceptance receipt; owner drafts and public publications require status readback. Suggestion submission also requires separate confirmation of the exact current ownership and license attestation shown in the preview.
Every published loop also includes a few useful parts:
- Use when explains the problem the loop is meant to solve.
- Prompt is the copy-ready instruction for your agent.
- Verify defines the evidence that proves the work succeeded.
- Steps show the feedback cycle in a more readable form.
- Notes call out practical limits, risks, or setup details.
- Related loops point to nearby workflows that may fit better.
Visit the Loop Library to browse published loops, copy one into your own workflow, or submit a loop that has worked well for you.
Loop Library is a Forward Future project and is available under the MIT License.
Notes for maintainers
Public loops are stored in the catalog database attached to the Cloudflare Worker. Publishing a reviewed loop does not require a GitHub commit or a static site deployment.
Copy loop-library/worker/examples/loop.json somewhere outside the repository,
fill in the record, and run:
LOOP_PUBLISH_TOKEN=... \
npm --prefix loop-library/worker run loop:publish -- /path/to/loop.jsonThe command validates the record and publishes the homepage row, detail page,
JSON/Markdown/plain-text catalogs, feed, and sitemap from the same database
write. Use --draft to save a non-public record or --archive to remove a
record from public responses without deleting its revision history.
The first database-backed release needs one import from the private migration bundle. Loop records and bootstrap data are intentionally not committed to GitHub:
LOOP_PUBLISH_TOKEN=... \
npm --prefix loop-library/worker run loops:import -- /private/path/bootstrap.jsonSet a long random LOOP_PUBLISH_TOKEN as a Worker secret. The catalog uses a
SQLite-backed Durable Object and keeps an append-only revision for every
publish. The reviewed bootstrap digest is enforced before the database can be
activated.
Create a private backup of the current database with:
LOOP_PUBLISH_TOKEN=... \
npm --prefix loop-library/worker run loops:export -- /private/path/catalog-backup.ndjsonRestore that snapshot only into a fresh, empty catalog database:
LOOP_PUBLISH_TOKEN=... \
npm --prefix loop-library/worker run loops:restore -- /private/path/catalog-backup.ndjsonBootstrap and backup files must be owner-only (chmod 600). Exports include
drafts, archived records, and complete revision history; keep them outside the
repository.
The current Git tree contains the site shell and rendering code, but no published loop records, generated loop pages, catalogs, feed, sitemap, or offline catalog fallback. The legacy catalog and source-attribution metadata were already public and intentionally remain in pre-migration Git history; this migration does not rewrite repository history or disrupt existing clones.
python3 -m http.server 4173 --directory loop-library/siteThen open http://localhost:4173.
npm ci --prefix loop-library/worker
node --check loop-library/site/script.js
node loop-library/scripts/check.mjs
npm --prefix loop-library/worker run check
python3 -m json.tool loop-library/site/.herenow/data.json >/dev/null
python3 -m json.tool loop-library/site/.herenow/proxy.json >/dev/null
python3 -m json.tool loop-library/scripts/seo-geo-query-benchmark.json >/dev/null
git diff --checkVoting is stored in a dedicated SQLite Durable Object. Reading totals is
public, but casting, changing, or removing a vote requires a GitHub login.
Set SESSION_SECRET and the GitHub OAuth client credentials as Worker
secrets; use loop-library/worker/.dev.vars.example for local variable names only. Register
the canonical callbacks shown in AGENTS.md, then deploy the Worker before the
site shell because the shell calls the new auth and vote routes.
The here.now proxy does not forward browser cookies or mutation Origin headers
and follows upstream redirects. The OAuth flow therefore uses an HMAC-signed,
browser-nonce-bound state value and a no-store callback bridge. The bridge saves
the signed session token in tab-scoped sessionStorage; session lookup and vote
writes send it only inside same-origin JSON request bodies.
Auth and proxy changes use a fail-closed staged rollout. Temporarily set
VOTING_UI_ENABLED=false while the Worker and proxy are deployed, then complete
a GitHub login, nonce-bound callback, session, vote, reload, and logout smoke
test on the canonical domain. Commit the value as the exact string true and
redeploy only the Worker after the smoke test passes; the already-published site
will reveal voting without another site publish.
Read AGENTS.md before editing loops or publishing the site. It contains the source-of-truth rules for database publishing, generated responses, form security, and clean-main deployments.
The legacy https://signals.forwardfuture.ai/* host is maintained by the
redirect-only Vercel project in
infra/signals-forwardfuture-ai-redirect/.
It permanently redirects paths to https://signals.forwardfuture.com/*.