Skip to content
Filip Stack
Documentation: fleet-sync

fleet-sync

Syncs a shared filip-stack baseline across linked T3 environments with per-machine exceptions.

Install
On its own
Runs
Picked when a request matches
Source
SKILL.md
npx skills add filipgutica/filip-stack -g -a claude-code codex -s fleet-sync -y

What it does

Fleet Sync applies one declared baseline through the same local updater on every enrolled machine. T3 Code keeps connection credentials and grants; your policy file owns the selections and exceptions. The result includes offline and unlinked machines.

When to use it

Install the skill, then ask:

$fleet-sync

The agent reads your roster from ~/.config/filip-stack/fleet.json, or from $XDG_CONFIG_HOME/filip-stack/fleet.json. On first use, identify the machines to enroll and choose the shared baseline. T3 may ask you to grant access to a machine. Valid grants and the saved selection are reused on later runs.

How it works

Each machine runs filip-stack sync --json with the shared policy and its environment ID, and saves that policy as its own. A differing saved policy is backed up first. Missing baseline skills and plugins are installed; existing extras remain. Per-machine agent targets and skill/plugin exclusions preserve exceptions. Existing scopes, enabled states, and explicit pins remain intact. Missing clients and local edits appear as incomplete components.

The workflow uses a separate clean main clone when a development checkout is busy. Guidance applied from an unmerged revision is retained until the selected source contains it or already matches all managed content. Other components can update while that guidance remains blocked. Custom Claude/Codex configuration directories require a guidance migration; the updater reports that component as blocked before writing default locations.

Use filip-stack sync --dry-run --json to preview one machine. A preview does not fetch upstream changes or prove that installed versions are current.

What you get

A result for every enrolled machine, with the shared policy digest, source commits, plugin versions, selections, and remaining conflicts or access problems. The command saves each machine's last result in ~/.local/state/filip-stack/last-sync.json, respecting XDG_STATE_HOME.

When chezmoi already transports shared Field Guide preferences, the fleet workflow pulls that source and merges its preferences through the Field Guide owner. Local unpublished guidance is preserved. Other dotfiles remain outside that preference refresh, and uninitialized setup phases are left untouched.

SKILL.md defines the complete workflow.