agentleFS
Sign inStart free
ryanmacleanryanmaclean/skills/jail-factory/SKILL.md

jail-factory

Skill1 starsChanged 50 days agoMITView on GitHub

Run bop card work inside disposable FreeBSD jails on fbryz3070 — golden ZFS dataset cloned per card, nullfs workdir r/w + skills r/o, destroy on exit. Use when the factory needs worker isolation, per-card credential scoping, or sandboxed agent execution on the BSD fleet.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Always umount -f ALL mounts (incl.

    line 52

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

jail-factory — disposable worker jails on fbryz3070

Host: root@fbryz3070.local (FreeBSD 15.0-p11, 128GB, NVIDIA 3070, pup+dd agent installed)

Layout

  • jails — 40G file-backed zpool at /jails.img (host is UFS; destroy pool + rm img to undo)
  • jails/golden@v2 — base snapshot: FreeBSD 15.0 + nu 0.116.1 (cargo-built — pkg binary needs FBSD_1.9), jq, git, pcre2, claude-code 2.1.283 (linuxulator), codex (cargo-built, compiling)
  • jails/f-<card> — per-card clone, mounted at /jails/f-<card>
  • /root/jail-factory/factory-run.nu — lifecycle: clone → nullfs+linprocfs/linsysfs/fdescfs mounts → jexec nu → destroy → propagate exit code
  • bop adapter: bop-bench-wt/adapters/jail.nu (stages workdir via scp, BOP_JAIL_CMD sets inner command)

Usage

ssh root@fbryz3070.local \
  'nu /root/jail-factory/factory-run.nu <card-id> <host-workdir> "<nu command>"'

Example:

nu factory-run.nu card-42 /work/card-42 "open /work/spec.md | print; exit 0"

Inside the jail

PathSourceMode
/workhost workdir (nullfs)r/w — card workspace
/skills/root/skills-factory (nullfs)r/o — skill corpus
/devdevfsminimal

ip4=inherit currently — tighten to a vnet/epair when credential scoping lands.

bop adapter contract

Wrap factory-run.nu in adapters/jail.nu conforming to <workdir> <prompt_file> <stdout_log> <stderr_log> [timeout] — exit 75 for rate-limit (EX_TEMPFAIL) so the provider chain rotates.

Gotchas

  • nushell + codex from pkg do NOT run on this base (built for FBSD_1.9, p11 ships FBSD_1.8) — both are cargo-built instead
  • claude runs under Linux emulation: jail needs linux=new + linprocfs/linsysfs mounts + fdescfs -o linrdlnk at /compat/linux/dev/fd, all mounted host-side (jail lacks mount perms for them)
  • jj is not packaged for FreeBSD — git only inside jails
  • Always umount -f ALL mounts (incl. linux fs) before zfs destroy or the dataset stays busy
  • freebsd-update refuses non-interactive runs; use freebsd-update cron (random sleep) or interactive

Roadmap

  • adapters/jail.nu in bop repo calling this over ssh
  • claude-code in golden via linuxulator
  • codex cargo build finishing (/tmp/codex-build.log on host)
  • Per-card env file (credential scoping — no inherited DD keys)
  • rctl limits: cpu/mem/pids per card
  • vnet networking instead of ip4=inherit
  • fleet-eval probe: verify "jail f-X ran" --probe "jls -j f-X" --expect empty

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free