superdesigndev/loopany

Última actividad de origen registrada: 7 de mayo de 2026 a las 07:57

Catálogo de SkillsMP actualizado: 29 de octubre de 2026 a las 18:30

skills recopiladas: 6

Estrellas en GitHub: 140
Forks en GitHub: 17

Instalar con un asistente de IA

Copia este prompt en el asistente de IA que estés usando.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/superdesigndev/loopany.

README del repositorio

Archivo de origen del README: README.md

Leer el README original

Mostrar README completo

# loopany <p align="center"> <img src="./screenshot.png" alt="loopany — self-improving companies" width="600" /> </p> **An agent skill for long-horizon work and self-improvement.** > **Where this comes from.** loopany is a side project from > [crewlet.io](https://crewlet.io) — the **self-improving company**: a > single AI agent that runs the company autonomously, learning from every > action it takes. loopany distills what we learned building that agent > into a portable, open substrate: the persistent memory and self-iteration > layer, extracted from the product and reusable by any agent harness. Most agents forget. They finish a task, the context window resets, and the next session starts blind — same mistakes, same blind spots, no memory of what worked. loopany is the persistent layer that fixes this: - **Long-cycle tasks** — the agent records what it did, what came back, and what's still open, so work that spans days or weeks survives across sessions. - **Self-learning** — every completed task writes an outcome. A reflect loop reads those outcomes, distills beliefs (`learning`) and proposes changes to its own skills (`skill-proposal`). The user accepts or rejects; accepted changes are diffed into the skill files and committed. The substrate is intentionally boring: markdown + frontmatter as the source of truth, append-only JSONL for the reference graph, no database. Any agent harness (Claude Code, Hermes, OpenClaw, Codex, custom) can attach this CLI and inherit a persistent, self-iterating brain. An optional local SQLite index is built on demand for search; it's a derived artifact and can be deleted at any time without data loss. ## Install Paste this into any agent harness (Claude Code, Hermes, Codex, …): ``` Retrieve and follow the instructions at: https://raw.githubusercontent.com/superdesigndev/loopany/main/INSTALL_FOR_AGENTS.md ``` The agent clones the repo, installs Bun + the CLI, initializes your workspace, runs a 5-phase onboarding conversation with you, and loads the skill library. You don't run any shell commands yourself. ## Learn more - [`CLAUDE.md`](./CLAUDE.md) — design philosophy, core concepts, MVP scope - [`INSTALL_FOR_AGENTS.md`](./INSTALL_FOR_AGENTS.md) — agent-facing install + command reference - [`ONBOARDING.md`](./ONBOARDING.md) — the 5-phase first-run script - [`skills/loopany-resolver/SKILL.md`](./skills/loopany-resolver/SKILL.md) — skill dispatcher (read this if you're an agent) - [`CHANGELOG.md`](./CHANGELOG.md) — release notes ## Testing ```bash # Unit + e2e tests (loopany CLI logic) bun test # Skill regression — installs all 5 skills into Claude Code, # runs 10 scenarios via `claude -p`, and verifies workspace artifacts. # Requires: claude CLI in PATH + active API key. ./test/skill-regression.sh # run all 10 scenarios ./test/skill-regression.sh 4 # run only scenario 4 ./test/skill-regression.sh --dry-run # show commands without calling claude ``` The skill regression covers: artifact creation (signal/task/note), capture quality gate, daily/weekly review, reflect threshold, signal→task upgrade, and kind routing. Each scenario spawns a fresh workspace, so runs are isolated. Expect ~5 min for a full pass (10 `claude -p` calls). ## Status `0.1.0` — usable for single-user dogfooding. Self-iteration loop (`reflect` skill writes proposals; the user applies accepted ones) is shipped; bundled domain packs (CRM / ads / content / metrics) are designed but not yet ported. ## Acknowledgments loopany stands on the shoulders of a few specific people's thinking: - **[GBrain](https://github.com/garrytan/gbrain)** by Garry Tan — the architectural sibling. Same overall shape (CLI + skills + markdown), different focus: GBrain answers *"what do I know"*, loopany answers *"what did I do, did it work, what should I change next."* - **[LLM Wiki](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)** by Andrej Karpathy — mental model for markdown-as-agent-memory and the latent-vs-deterministic division of labor.

Cargando el README del repositorio…

Skills en este repositorio

Otras ocupaciones informáticas Desarrolladores de software

2 categorías ocupacionales · 100% clasificado

skill

ocupación

descripción

actualizado

loopany-core

Otras ocupaciones informáticas

Artifact lifecycle for loopany — the agent brain. Use when creating, modifying, or querying any artifact (task, signal, learning, skill-proposal, mission, person, note, brief, journal). Read this SKILL.md first for routing, then the relevant kind's Playbook.…