← ClaudeAtlas

autonomy-budgetlisted

Hand a multi-day job to your agent and walk away — for real. You set the rails once (what it can touch, what it can never touch, how long it runs, what pulls it back to you), then it works for days on its own and comes back only when something genuinely needs a human. It saves its place so a crash or a closed laptop doesn't cost the work, and it leaves a log that proves it stayed inside the rails. Use when you want days of throughput without babysitting — a batch audit of every client site, a multi-site refresh, a long overnight-into-the-next-day job — when you're afraid to leave an agent running unwatched because you can't be sure what it'll touch, or when you're on Level 9 and want the "set the rails and leave" step done right.
search-atlas-group/amm-founding-circle · ★ 0 · AI & Automation · score 72
Install: claude install-skill search-atlas-group/amm-founding-circle
# autonomy-budget **The problem this solves:** you've got an agent that can do real work, but you're still chained to it. A big job — audit all 40 client sites, refresh every location page, run a week of analysis — means a week of you starting runs, checking in, restarting when it dies, and hoping it didn't touch something it shouldn't. You can't leave it alone for days because you can't *prove* what it would do while you're gone. So the biggest jobs still cost your whole week. `autonomy-budget` fixes that. It's not "let the agent do whatever it wants" — it's the opposite. You give the agent an **autonomy budget**: a fixed set of rails it runs inside for a fixed stretch of time. Inside the rails it moves fast and decides on its own; it can't step outside them; and the handful of things that genuinely need you are the *only* things that pull it back. You set the budget once, hand it a multi-day job, and close the laptop. The offense is days of work done while you were away — the rails are the seatbelt that lets you drive that fast without blowing up a client account. > This is the **multi-day** version of always-on. Its single-night sibling is `night-shift` — the bounded-overnight contract (time box, worker cap, read-only-by-default, failure ledger). Read `night-shift` once; it's the reason a run you're not watching is trustworthy. `autonomy-budget` scales that contract across *days* and composes three L5 pieces to make a multi-day run survivable: `goal-mode` gives it a fin