Shell Sync
Skill0 starsChanged 4 months agoMITView on GitHub
Set up, repair, upgrade, and optimize the sync between your zsh and fish shells, and keep PATH healthy. Use to SET UP a canonical→mirror sync (regenerate one shell's PATH/env/aliases/functions and prompt from the other so both behave the same, including a shared starship prompt), REPAIR PATH and divergence (dead entries, installed-but-not-on-PATH tools, a tool or prompt resolving differently per shell, startup errors), UPGRADE the mirror after the canonical shell changes (re-propagate new aliases/vars/PATH/prompt), or OPTIMIZE a working setup (deduplicate, fix ordering/precedence, prove parity). Every change is verified by resolving the environment in both shells and comparing.
In this repository12 other files it gives its agents.
- Brew DoctorSkill
- Credential StoreSkill
- Dev EnvironmentSkill
- DotfilesSkill
- Ghostty ConfigSkill
- Git SetupSkill
- Identity ProfilesSkill
- macOS DefaultsSkill
- Neovim ConfigSkill
- Runtime VersionsSkill
- SSH ConfigSkill
- Terminal ThemeSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Keep zsh and fish consistent, and keep PATH healthy.
- Distinctive: line 27
No mirror file exists yet → setup.
- Very specific: line 67
Instead run the canonical shell and read the resolved state with scripts/dump_env.py, then emit it for the mirror:
- Very specific: line 104
Verify in both shells: After syncing, resolve PATH/env/aliases in *both* and diff them — see reference/verification.md.
- Very specific: line 111
Fix the canonical first: If PATH is messy, run the repair workflow on the canonical shell, then re-sync (upgrade.md) — don't propagate junk …
- Very specific: line 117
Read ~/.config/devenv/config.toml before prompting.
≈ “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.
Files
Shell sync
Keep zsh and fish consistent, and keep PATH healthy. Route to a workflow:
| The user wants to… | Workflow |
| Establish the sync — make both shells behave the same (env, PATH, aliases, functions) | reference/setup.md |
| Fix PATH or divergence — dead entries, a tool not found, a tool differing per shell | reference/repair.md |
| Re-sync the mirror after changing the canonical shell (new alias/var/PATH) | reference/upgrade.md |
| Tighten a working setup — dedupe, fix ordering/precedence, prove parity | reference/optimize.md |
Sync model is canonical → mirror: one shell is the source of truth; the other is regenerated
to match. Default canonical = the login shell ($SHELL); confirm with the user.
No workflow named? (a bare /shell-sync) — don't guess one. Triage first: run Discovery,
then scripts/shell_diff.py and scripts/mirror_drift.py, classify the findings, and route:
mirror_drift.pyreports drift (drift_*/stale_mtime) → upgrade (re-sync the mirror).shell_diff.pyshows areviewdivergence or astartup … issues→ repair.- No mirror file exists yet → setup.
- Everything
benign/in_sync yes/startup clean → report parity and stop — there's nothing to do.
Ignore benign (…) divergences from shell_diff.py (system path_helper / brew vendor
activation) — they're expected noise, not work. Only a review row is a real divergence.
The scripts (call these, don't re-derive them)
The mechanical commands live in scripts/ as tested, clean-environment helpers.
Call them by name — the inline commands below and in the reference docs is under-the-hood explanation,
not what you retype each session. All are read-only; none writes a config file (repairs/mirrors print
a plan to stdout for you to apply).
| Need | Script |
| Resolve a shell's real PATH / exports / aliases / functions | scripts/dump_env.py <zsh|fish> [section] |
Audit PATH — dupes, dead entries, installed-but-not-on-PATH (--plan = cleaned PATH) | scripts/path_doctor.py [zsh|fish] [--plan] |
| Diff zsh vs fish — PATH set, startup cleanliness, tool reachability | scripts/shell_diff.py [tool …] |
| Emit the proposed fish mirror file (zsh→fish: PATH/env/aliases + starship prompt init) to stdout | scripts/mirror_plan.py |
| Check the installed mirror is still current vs the canonical (drift / staleness) | scripts/mirror_drift.py [mirror-file] |
Discovery (always run first)
echo "login shell: $SHELL"; command -v zsh fish # which shells existLocate config files (only those that exist):
- zsh:
~/.zshenv,~/.zprofile,~/.zshrc,~/.zlogin - fish:
~/.config/fish/config.fish,~/.config/fish/conf.d/*.fish,~/.config/fish/functions/,~/.config/fish/fish_variables(universal vars incl.fish_user_paths) - macOS system PATH:
/etc/paths,/etc/paths.d/*(applied bypath_helperin/etc/zprofile) - prompt:
command -v starshipand~/.config/starship.toml— starship is cross-shell, so its single config is shared and only its per-shell init line is mirrored (see below).
Establish canonical (source of truth) and mirror (regenerated) — confirm direction before writing anything.
Core method: extract resolved values, don't parse script
Shell configs contain eval, source, and conditionals — do not statically translate them.
Instead run the canonical shell and read the resolved state with scripts/dump_env.py, then
emit it for the mirror:
scripts/dump_env.py zsh path # resolved PATH entries, one per line
scripts/dump_env.py zsh exports # exported env vars, NAME=value
scripts/dump_env.py zsh aliases # aliases, name=value
scripts/dump_env.py zsh functions # user function namesdump_env.py already runs the shell in a clean environment (env -i … zsh -l -i -c …) so a
shell launched inside another session can't pollute the result with the parent's PATH.
<details><summary>Under the hood (what dump_env.py runs)</summary>zsh -l -i -c 'printf "%s\n" $path' # resolved PATH entries (zsh $path array)
zsh -l -i -c 'env' # exported env vars, NAME=value
zsh -i -c 'alias' # aliases
zsh -i -c 'print -l ${(k)functions}' # user function names</details>For a mirror, drop system defaults/shell-internal vars (mirror_plan.py applies a denylist; or
compute a delta vs a bare shell, env -i zsh -fc 'env'). Translation rules (zsh↔fish for
PATH/env/alias/function/abbr): reference/translation.md.
Core principles
- Regenerate into the marked block; never hand-merge outside it. The mirror's synced content
lives in one marked, auto-generated file (e.g. fish
conf.d/00-shell-sync.fish) the workflow overwrites idempotently — never splice into a file the user hand-edits.mirror_plan.pyemits that block, but treat it as a draft to review, not to install blindly: its filters are good but not omniscient (a new tool that injects env likemise/brew, an alias it mis-classifies). When the generator's output would regress a good curated file, a surgical edit inside the markers — or a one-line diff-and-patch guided bymirror_drift.py— is the correct move, not a full overwrite. Order of preference: regenerate → marked-block edit → (never) hand-merge outside the markers. - Verify in both shells. After syncing, resolve PATH/env/aliases in both and diff them — see reference/verification.md. This is the golden rule: prove it, don't assume.
- Back up before overwriting any config file (timestamped copy).
- Be honest about limits. Functions and complex aliases translate poorly across shells — auto-translate the simple ones, and list what needs manual review rather than emitting broken code.
- Fix the canonical first. If PATH is messy, run the repair workflow on the canonical shell,
then re-sync (upgrade.md) — don't propagate junk (
/path/to/pip, duplicates) into the other shell.
Settings
Read ~/.config/devenv/config.toml before prompting; if a [shell-sync] section exists, use it as
defaults (precedence: explicit answer this session > config.toml > ask). Had to ask? Offer to save
the answer back. Managed by the devenv skill. Keys honored:
canonical— source-of-truth shell (skips the "which is canonical?" prompt).mirror— the shell regenerated to match.sync— what to keep in sync: any ofpath,env,aliases,functions,prompt(the starship init line; the sharedstarship.tomlis tracked by thedotfilesskill).
Safety
- Confirm sync direction before writing. Overwriting the wrong shell's config loses work.
fish_add_pathandset -gx PATHchange PATH ordering — regenerate the whole managed block so ordering is predictable, don't append piecemeal.- Removing PATH entries can hide tools — show what would be removed and why, and confirm.
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
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.

