From abatilo-core
Work autonomously on a task until it is complete or the human stops you, without pausing to ask permission between steps. Use when user says "/auto-loop", "loop forever", "keep going", "run overnight", "don't stop", "run autonomously", "work until it's done", "finish this without me", or when the user is leaving and wants work to continue unattended. Applies to both open-ended iteration (optimization, experimentation, search) and finite tasks that need to be driven to completion (migrations, refactors, implementing a feature, fixing a list of issues).
npx claudepluginhub abatilo/vimrc --plugin abatilo-coreThis skill uses the workspace's default tool permissions.
You are about to enter autonomous mode. You will work continuously until the task is complete or the human manually stops you. This skill turns you into an autonomous agent that drives a task to completion without waiting for permission at each step.
Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).
Generates original PNG/PDF visual art via design philosophy manifestos for posters, graphics, and static designs on user request.
Share bugs, ideas, or general feedback.
You are about to enter autonomous mode. You will work continuously until the task is complete or the human manually stops you. This skill turns you into an autonomous agent that drives a task to completion without waiting for permission at each step.
Work with the user to establish:
If the conversation already makes these clear, confirm your understanding and begin. Do not over-plan — start the loop quickly.
results.md to track progress: what you did each step, what happened, and whether you kept or discarded it.LOOP UNTIL DONE:
results.md — what you did, the outcome, and whether you kept or discarded it.Once the loop has begun, do NOT pause to ask the human if you should continue. Do NOT ask "should I keep going?" or "is this a good stopping point?". The human might be asleep, away from the computer, or otherwise occupied — they expect you to keep working until the task is done or they manually stop you. You are autonomous.
There are exactly two reasons to stop:
Everything else — errors, failed attempts, running out of obvious ideas — is NOT a reason to stop. It is a reason to adapt.
If you run out of ideas, think harder:
results.md for patterns — what kinds of changes helped vs hurt?mcp__codex__codex) with a summary of the task, what you have tried so far, and what worked vs failed. Ask it for the next set of ideas or a different angle of attack. Use threaded replies (mcp__codex__codex-reply) to refine its suggestions before trying them. This is your brainstorming partner — use it before giving up.Crashes: Use your judgment. If the error is trivial (a typo, a missing import, an off-by-one), fix it and re-run. If the idea itself is broken, log it as a failure and move on to the next idea. Do not spend more than 2-3 attempts fixing the same crash.
Timeouts: If an iteration takes far longer than expected, kill it and treat it as a failure. Discard and move on.
Stuck: If you are not making progress, step back. Re-read your results.md. Look for patterns. Try a completely different direction. Stalling is not a reason to stop — it is a reason to think differently.
Each iteration generates output. To avoid filling your context with noise:
results.md concise — one line per iteration.results.md to remind yourself of the trajectory.When you finish (goal achieved) or when the human returns, they will want to see:
results.md showing everything you tried and what worked.