agentleFS
Sign inStart free
khivikhivi/cockpit/.claude/skills/cockpit-dev/SKILL.md

cockpit-dev

Skill0 starsChanged 41 days agoMITView on GitHub

Run this worktree's cockpit build against a throwaway sandbox, so it can't fight the installed daemon or touch real worktrees. TRIGGER on "run cockpit", "try it in the TUI", "does this render right". Not for pytest or `cockpit setup`.

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:

    Run ./dev.sh from the repo root.

    line 11

≈ “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

cockpit-dev

Run ./dev.sh from the repo root. It seeds .cockpit-dev/ and execs this worktree's build against it.

./dev.sh                 # your real repos + a copy of your real PR cache, read-only
./dev.sh --empty         # no repos — layout and keybinding work
./dev.sh -- nudge list   # any other subcommand, same sandbox

watch is a Textual TUI and needs a terminal. If you're in a session without one, tell the user to run it themselves — suggest they type ! ./dev.sh so the output lands in the conversation, or run it in their own terminal.

What the sandbox does and does not cover

./dev.sh isolates five things. Read dev.sh's header comment for why each one is load-bearing; the short version is that missing any single one reaches real state:

IsolatedCovers
COCKPIT_HOMEconfig.json, PR cache
COCKPIT_RUNTIME_DIRcockpit.pid, close-request queue — a separate variable, because these are machine-local and deliberately do not follow COCKPIT_HOME
TMPDIRstatusline/starship flat cells (they live in $TMPDIR/cockpit-cache, not under COCKPIT_HOME)
tool: noneevery cmux write — spawn, close, rename, set-color, workspace-group, send
--dryteardown, autoclose, git branch -D, tracker writes, spawn, nudge, the fast tick's cmux rename/recolour, and the row keys that reach outside (n/f/h/a, which refuse)

Two things it does not cover, and you should say so rather than imply full coverage:

  1. Every cmux-facing feature is inert. Sidebar folds, f/focus, a/ask, d/diff, colours, workspace groups — all no-ops under tool: none. Testing a change to any of those needs a real cmux, and therefore real risk. The sandbox is right for the table, cells, config, prompts, and the reconcile cycle's decisions.
  2. gh reads still happen on the slow tick and spend rate limit.

Never run cockpit setup from a worktree

dev.sh refuses it, exit 2. Setup bakes sys.executable — here .venv/bin/python — into ~/.claude/settings.json and ~/.config/starship.toml, both outside the sandbox. When the worktree is removed, the interpreter goes with it and the user's statusline dies. That's the "footer disappeared" bug in AGENTS.md's {python} pin invariant. The fix is re-running the brew-installed cockpit setup, so don't work around the refusal.

Reading the result

Under --dry the daemon prints its decisions with a [dry] prefix instead of acting. That output is the point: it's how you check that a change decides correctly without letting it act. The table renders from the seeded cache, since --dry suppresses cache writes.

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