Pre-built agent setups you can fork into a fresh Firecracker microVM. Forks capture the original creator's /persist — config, conversation memory, custom skills, datasets. Forking creates a new paid microVM ($25+/mo, 30-day money-back).
116 public templates · Collections · GTM tools · Read the guide
Claude Code
Internal battle cards where every competitor claim is dated and sourced — and the 'when they win' section is mandatory. Never publishes externally.
Claude Code
Tests every Linear bug against the four things that make it reproducible, then drafts the exact questions for whatever is missing. Never closes an issue.
Claude Code
Drafts eBay listings that fit the 80-char title cap, fill every item specific, and price from sold comps — not asking prices. Never lists.
Claude Code
Finds the Sponsored Products spend that never converted, then drafts harvest, negate and bid-down changes with the arithmetic shown. Never changes bids.
Claude Code
A three-email recovery sequence with no discount in email one — reminder, objection, last call. It won't invent urgency, and it never sends.
Claude Code
Finds the docs your codebase contradicts — dead commands, moved paths, broken links — with sourced corrections. Age isn't rot, and it never edits your wiki.
Claude Code
True margin from landed cost — not markup, not invoice cost — plus price scenarios with the break-even volume change. Never changes prices.
Claude Code
Drafts PayPal dispute responses sorted by deadline, with every claim matched to an artifact and the unprovable ones listed as gaps. Never submits.
Claude Code
Tracks competitor prices on landed cost with variant and stock context, and reports movement against recorded history. Never changes your prices.
Claude Code
Days-of-cover reorder brief from your sales and stock exports — and it flags the SKUs whose velocity a stockout quietly suppressed. Never orders.
Claude Code
Reconciles bank, processor, AR and AP exports into a close packet with tie-outs shown. Unexplained variance is a finding — it never plugs. Never posts entries.
Claude Code
Ranks Datadog monitors by action rate, not volume — then proposes specific thresholds with the percentile behind them. Never mutes a monitor.
Claude Code
Sorts a Dependabot backlog into routine, needs-migration and security — then drafts the groups and ignore rules that stop it regenerating. Never merges.
Claude Code
Triages GitHub secret-scanning alerts by blast radius into a rotate-first runbook. Never closes an alert, never prints the secret.
Claude Code
Ranks accounts by change from their own baseline into a HIGH/MEDIUM/WATCH triage list with the question to ask. No fake probabilities. Never contacts customers.
Pi
Control flow inside Pi: TypeScript graphs with y/n branches, implement-verify loops, seven seeded workflows, live graph widget. Bring your model key.
Claude Code
Draft Shopify support replies and review responses on your own read-only token — no per-resolution meter. It writes drafts. You send them.
Claude Code
Fork a Claude Code box with the Loopany Reddit karma loop: wiki, filters, ledger, OpenCLI. Draft-first. Your account.
Hermes Agent
Cheat-code demand loop: landing + ads (Higgsfield), de-anon, ICP, MillionVerifier, Instantly drafts, Plunk nurture drips, inbox demos. RB2B optional.
Claude Code
Error cluster → timeline, suspect commits, draft postmortem. You page. Never auto-mute or auto-deploy.
Claude Code
Schema/migration diffs with expand/contract safety, lock risk, and rollback notes. You still ship.
Claude Code
When main goes red, draft the smallest green fix PR. Human merge gate. Never force-push main.
Claude Code
Sequence, contracts, and data model deltas before sprawl. Lights on for long-lived decisions. Pair with program design.
Claude Code
Directional maintainability smells on a PR — complexity, duplication, shotgun risk. Draft review only. Tests green ≠ maintainable.
Claude Code
Rewrite horizontal migrations→API→UI plans into tracer-bullet slices you can curl and review every 100–200 lines.
Claude Code
Problem, success metric, and rough HTML mockups before any code. Mockups are cheaper than slop PRs.
Claude Code
Large PRs get a logical diff tree — ordered by change intent, not alpha. Review energy heuristic. You still read the code.
Claude Code
Call-stack trees, file-tree diffs, and types before implementation. Architecture alone is not enough. You approve design first.
Claude Code
Research → plan → implement skills with intentional compaction. Human gates on research and plan. Draft only — you still merge.
Hermes Agent
Ads dashboard with a budget gate: drafts hooks and CTAs to an Obsidian board you own, sets break-even ROAS before a dollar of spend. One channel, mastered.
Ready to fork
Pick a public template, fork it to a fresh Firecracker microVM, add your model keys, and put it to work. 30-day money-back · no free trial · from $25/mo.