
chezmoi
Manage dotfiles with chezmoi — the source-state dotfile manager that uses filename attributes (dot, private, encrypted, runonce), Go templates, and password-manager integration to keep one set of dotfiles working across many machines. Use when the user says "set up dotfiles", "manage my dotfiles", "bootstrap a new machine", "chezmoi", "encrypt my SSH key in dotfiles", "template dotfiles per OS", "add secrets to dotfiles", "what should be in my .chezmoi.toml.tmpl", or asks how to share configs across macOS / Linux / servers. Also trigger when the user is staring at a fresh machine and wants the one-liner that gets them home, or when they're about to commit something sensitive into a dotfiles repo. Do NOT use for stow / yadm / rcm or other dotfile managers, generic git repo setup, or password-manager setup unrelated to dotfiles.
- Installs
- 58
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/paulnsorensen/skillz-that-grillz --skill chezmoi- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads chezmoi automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
chezmoi
Wrap chezmoi — manage dotfiles across machines using source-state semantics, Go templates, and at-rest encryption.
When to use
- Bootstrapping dotfiles on a new machine
- Adding a new file to existing chezmoi-managed dotfiles
- Templating a dotfile per-OS, per-host, or per-role (work vs. personal)
- Adding secrets via 1Password / Bitwarden / age / gpg
- Writing a
run_once_install script - Diagnosing "chezmoi apply changed something I didn't expect"
- Designing the
.chezmoi.toml.tmplconfig template for a fleet of machines
Mental model
chezmoi maps a source directory (~/.local/share/chezmoi, a git repo) to
a target directory (the user's home). Source filenames carry attribute
prefixes that determine target name and behavior. Source files with a .tmpl suffix are rendered as Go text/template and
written to their target paths. Files under .chezmoitemplates/ are reusable
template fragments included by other templates — they do not produce target
files on their own. A
config file at ~/.config/chezmoi/chezmoi.toml (typically generated from
.chezmoi.toml.tmpl on first init) holds per-machine variables.
The two reasons people get into trouble:
- They edit files directly under
~/.local/share/chezmoi/instead of usingchezmoi edit $TARGET. This breaks templates and decrypt round-trips. - They commit secrets in plaintext "because the repo is private". Repos leak, forks leak, history is forever. Use a secret backend.
Where the config file lives
Two files, easy to confuse. Both can be in toml, yaml, json, or
jsonc (auto-detected by extension; two formats present at once is an
error, not "first one wins"):
| File | Path | Who writes it |
|---|---|---|
chezmoi.<fmt> (the config) | $XDG_CONFIG_HOME/chezmoi/chezmoi.<fmt>, default ~/.config/chezmoi/chezmoi.<fmt> (.toml most common). NOT in the source repo. | Generated by chezmoi init (or apply --init) from the template below. Per-machine. Holds secrets, API tokens, machine-specific settings — never commit. |
.chezmoi.<fmt>.tmpl (the config template) | Source-state root: ~/.local/share/chezmoi/.chezmoi.<fmt>.tmpl (.toml most common). Committed to the dotfiles repo. | You write it. Uses promptBoolOnce / promptChoiceOnce / promptStringOnce so a fresh machine generates the right config on first init. |
Override locations with --config <path> (use --config-format <fmt> if
the extension is unusual). Look up effective values with
chezmoi cat-config and chezmoi dump-config.
Protocol
1. Detect what the user wants
| User says | Go to |
|---|---|
| "set up chezmoi from scratch" / "new machine" | §2 Bootstrap |
| "add this file to my dotfiles" | §3 Add a file |
| "make this dotfile depend on OS / host" | references/templating.md |
| "encrypt this" / "store this token" | references/secrets.md |
| "run a script on first apply" | references/scripts.md |
| "chezmoi did something weird" | references/pitfalls.md |
| "design my full setup" | references/bootstrap.md |
2. Bootstrap (new machine)
The full one-liner that clones, renders templates, and applies in one shot:
sh -c "$(curl -fsLS get.chezmoi.io)" -- init --apply $GITHUB_USERNAME
For a fully fleshed .chezmoi.toml.tmpl recipe with promptBoolOnce for
work-vs-personal-vs-headless, see references/bootstrap.md.
3. Add a file
chezmoi add ~/.zshrc # plain copy
chezmoi add --template ~/.gitconfig # add as a template (.tmpl suffix added)
chezmoi add --encrypt ~/.ssh/config # encrypted at rest
chezmoi chattr +template ~/.zshrc # turn an existing managed file into a template
chezmoi edit ~/.zshrc # edit the SOURCE — handles decrypt + .tmpl correctly
Then the safe-apply ritual (§5).
4. File-naming reference
The table users get wrong most often. Prefixes must appear in a specific
order, and the allowed prefix set depends on the target type —
/reference/source-state-attributes/ is the canonical source.
| Prefix | Effect on target |
|---|---|
dot_ | Add leading . (dot_zshrc → ~/.zshrc) |
private_ | Strip group + world permissions (mode 0600 / 0700) |
readonly_ | Strip all write permission bits |
executable_ | Add executable bit |
empty_ | Allow empty file (default would remove it) |
encrypted_ | Decrypt source via age/gpg before writing target |
symlink_ | Create a symlink instead of a regular file (file form, target is symlink target) |
exact_ (dirs) | Delete anything in target dir not present in source |
create_ | Create target only if missing — never overwrite |
modify_ | Treat source as a script (or template, see below) that produces target contents |
run_ | Treat source as a script to execute (not a target file) |
before_ / after_ | Order scripts relative to applying files |
once_ (scripts) | Run only when contents have not been run successfully before |
onchange_ (scripts) | Run only when contents change for THIS filename |
remove_ | Remove the target if it exists |
external_ (dirs) | Mark a directory as an external — pairs with .chezmoiexternal.<fmt> / .chezmoiexternals/; chezmoi ignores attributes on child entries |
literal_ (anywhere) | Stop parsing further attributes from this point |
Allowed prefix order per target type (the order is rigid; rearranging is a parse error):
| Target type | Allowed prefixes (in order) | Optional suffix |
|---|---|---|
| Directory | remove_, external_, exact_, private_, readonly_, dot_ | — |
| Regular file | encrypted_, private_, readonly_, empty_, executable_, dot_ | .tmpl |
| Create file | create_, encrypted_, private_, readonly_, empty_, executable_, dot_ | .tmpl |
| Modify file | modify_, encrypted_, private_, readonly_, executable_, dot_ | .tmpl |
| Remove file | remove_, dot_ | — |
| Script | run_, once_ or onchange_, before_ or after_ | .tmpl |
| Symlink | symlink_, dot_ | .tmpl |
Suffix .tmpl makes the file a template. .age / .asc are stripped
automatically when the corresponding encryption is configured.
Modify file detail. A modify_ source is a script by default — it reads
the current target on stdin and writes the new target to stdout. If the
file contains the string chezmoi:modify-template, lines containing that
marker are stripped and the rest is interpreted as a template; the existing
target is exposed in .chezmoi.stdin. Use the template form when you want
to surgically edit a structured file (JSON, TOML, YAML); use the script
form when you need a real interpreter (jq, awk, sed).
5. Safe-apply ritual
Before any apply that touches important files, run the inspection trio:
chezmoi status # one line per changed file
chezmoi diff # unified diff of pending changes
chezmoi apply --dry-run -v # full plan, including scripts that would run
chezmoi apply # only after the above looks right
For routine pulls on a machine you trust:
chezmoi update # = git pull --autostash --rebase + apply
chezmoi git pull -- --autostash --rebase && chezmoi diff # preview only — no apply
6. Edit safely
chezmoi edit ~/.zshrc # opens source in $EDITOR, handles .tmpl + encryption
chezmoi cd # subshell in the source directory (for git operations)
chezmoi re-add # refresh source from current target after editing target
chezmoi merge ~/.zshrc # 3-way merge target ↔ source ↔ destination
chezmoi merge-all # 3-way merge every file whose actual ≠ target (add --init to re-render config first)
Editor resolution: edit.command → $VISUAL → $EDITOR → vi (or
notepad.exe on Windows). chezmoi prints a warning if the editor returns
in < edit.minDuration (default 1s) — that catches the "I opened the
file but didn't actually edit it" case. Set to 0 to silence.
Never open ~/.local/share/chezmoi/encrypted_dot_ssh/... or *.tmpl files
directly in your editor — the file you see isn't the file chezmoi sees.
7. Application order (mental model)
chezmoi apply is deterministic. Each run does:
- Read source state, read destination state, compute target state.
- Run
run_before_scripts in ASCII order of target name. - Update entries in ASCII order of target name. Directories (including directories materialized by externals) are written before the files they contain. Externals are fetched/expanded during this phase.
- Run
run_after_scripts in ASCII order.
Implications:
- A
run_before_script must not read a file that an external will provide — externals are applied in phase 3, not before. - Target names are compared with all attribute prefixes stripped. Given
create_alphaandmodify_dot_beta,.betais updated beforealphabecause.beta<alphain ASCII. - chezmoi assumes source and destination don't change during a run. A
run_before_script that writes into either violates that assumption — behavior is undefined.
8. Partial-file management
When chezmoi shouldn't own the whole file, pick a pattern based on who the file belongs to:
modify_ — user owns the file, chezmoi stamps a block. The source
reads the existing target on stdin and writes the new target on stdout.
Script form, modify_dot_zshrc:
#!/usr/bin/env bash
set -euo pipefail
# Strip any prior chezmoi block from the existing file …
awk '/# chezmoi:begin/{s=1;next} /# chezmoi:end/{s=0;next} !s'
# … then append the current managed block.
cat <<'EOF'
# chezmoi:begin
export EDITOR=nvim
alias g=git
# chezmoi:end
EOF
Template form (for structured files), modify_dot_config_private_app.yaml.tmpl:
{{- /* chezmoi:modify-template */ -}}
{{- $cfg := dict -}}
{{- if .chezmoi.stdin -}}{{- $cfg = .chezmoi.stdin | fromYaml -}}{{- end -}}
{{- $_ := set $cfg "telemetry" false -}}
{{- $_ := set $cfg "user" (dict "email" .email) -}}
{{ toYaml $cfg }}
The if .chezmoi.stdin guard handles the initial-apply case where the
target doesn't exist yet. Use fromJsonc / fromToml / fromYaml +
toPrettyJson / toToml / toYaml for the round-trip.
.chezmoitemplates/ — chezmoi owns the file, fragments are shared.
Reusable blocks live under .chezmoitemplates/, pulled into regular
templates with {{ template "name" . }} (or includeTemplate "name" .
when you need to pipe the result through a function):
.chezmoitemplates/zsh-prompt # one file
dot_zshrc.tmpl # {{ template "zsh-prompt" . }}
dot_config/fish/config.fish.tmpl # {{ template "zsh-prompt" . }} too
Choose the pattern by ownership. modify_ when the user can freely
edit the file outside the managed block. .chezmoitemplates/ when
chezmoi renders the whole file but you want to DRY shared sections.
Mixing both in one target is usually a smell.
Hard rules
These are non-negotiable. Each one corresponds to a real foot-gun:
- Never commit plaintext secrets, even to a private repo. Repos leak,
get forked, get logged, get backed up. Pick a backend from
references/secrets.md. - Never edit encrypted or templated source files directly. Use
chezmoi edit $TARGETso the decrypt and template round-trips happen. - Always inspect before applying on shared/important files —
chezmoi difforchezmoi apply --dry-run -v.exact_on a directory deletes unmanaged entries; you want to see that coming. prompt*template functions belong only in the config-file template (.chezmoi.toml.tmpl). Using them in regular dotfile templates causes a prompt on every apply / diff / status, which defeats automation.
Secrets decision tree (summary)
Full details in references/secrets.md. Pick exactly one per repo.
- Public repo + cloud password manager → 1Password CLI / Bitwarden template functions. Secrets fetched at apply time, never on disk.
- Private repo + offline-friendly → age (
encrypted_prefix). Identity at~/.config/chezmoi/key.txt, lives outside the repo. - Sharing across a team → SOPS-encrypted
.chezmoidata/secrets.yaml. - Air-gapped / first-bootstrap → gitignored plaintext
.chezmoidata/local.yaml(transitional, not a long-term answer).
Mixing backends multiplies moving parts without making anything more secure.
Templating quick reference
Full details in references/templating.md.
{{- if eq .chezmoi.os "darwin" }}
export HOMEBREW_PREFIX="/opt/homebrew"
{{- else if eq .chezmoi.os "linux" }}
export PATH="/usr/local/bin:$PATH"
{{- end }}
{{- if and (eq .chezmoi.os "darwin") (lookPath "brew") }}
# brew is installed
{{- end }}
email = {{ .email | quote }}
Built-in template variables include .chezmoi.os, .chezmoi.arch,
.chezmoi.hostname, .chezmoi.username, .chezmoi.homeDir,
.chezmoi.sourceDir. User-defined data lives under [data] in
chezmoi.toml, or any .chezmoidata/*.{toml,yaml,json} file.
chezmoi runs templates with Go's missingkey=error option, so a typo
like {{ .gitub.user }} fails the apply instead of rendering empty.
Override via [template] options = ["missingkey=zero"] only if you have
a concrete reason — losing the typo guard is rarely worth it.
For symlink targets, leading/trailing whitespace in the rendered result
is stripped, and an empty result removes the symlink. For file targets,
an empty render removes the file (use the empty_ prefix to keep it).
Special files and directories
All entries in the source dir beginning with . are ignored except the
specials below. They are evaluated in this order on every operation:
.chezmoiroot → .chezmoi.<fmt>.tmpl → data files → .chezmoitemplates/
→ .chezmoiignore → .chezmoiremove → externals → .chezmoiversion.
| Path | Role |
|---|---|
.chezmoiroot | Single-line file pointing at a sub-directory as the source root |
.chezmoi.<fmt>.tmpl | Config-file template used on first chezmoi init (and on --init) |
.chezmoidata.<fmt> / .chezmoidata/ | Data merged into the template . namespace before any template renders |
.chezmoitemplates/ | Reusable template fragments ({{ template "name" . }}) — no target files |
.chezmoiscripts/ | Scripts not bound to a target file; honor run_before_ / run_after_ |
.chezmoiignore | Per-target ignore list — source-relative, templatable |
.chezmoiremove | Targets to remove on apply — templatable |
.chezmoiexternal.<fmt> / .chezmoiexternals/ | Pull files/archives from URLs at apply time |
.chezmoiversion | Min chezmoi version required by this source repo |
.chezmoiignore matches source-relative paths (e.g. dot_zshrc), not
target paths. People get this wrong constantly.
<fmt> means toml, yaml, json, or jsonc. The directory forms
(.chezmoidata/, .chezmoiexternals/) merge their contents in lexical
order with any same-prefix file forms — useful for splitting big config
across files or composing externals per role.
What you don't do
- Write a competing dotfile manager or shim around
chezmoi apply - Edit files under
~/.local/share/chezmoi/directly when they are templates or encrypted - Commit a
.env/secrets.yamlto a "private" repo - Suggest
chezmoi applywithout first runningchezmoi diffon a shared machine - Use
prompt*template functions outside of.chezmoi.toml.tmpl - Mix multiple secret backends (age + 1Password + SOPS) in the same repo without a clear, written reason
Gotchas
fork/exec ...: exec format errorrunning a templated script means a newline before#!. Add-inside the closing}}on the first template line to suppress it.fork/exec ...: permission deniedfor scripts means$TMPDIRis mountednoexec. SetscriptTempDirinchezmoi.toml.- Windows line endings (CRLF) in templates produce broken shell scripts on
Linux/macOS. Add
* text eol=lfto a.gitattributesin the dotfiles repo, or setcore.autocrlf = input. chezmoi initon a new machine before.chezmoi.toml.tmplexists silently skips per-machine prompts. Land that template before adding host-specific files.chezmoi doctoris the first command to run when something is off — it reports missing dependencies (op,age, etc.), broken paths, and shell integration issues in one pass.- Context7 may be unavailable. The chezmoi CLI is self-documenting via
chezmoi helpandchezmoi help <command>; fall back to that. - Two config files (e.g. both
chezmoi.tomlandchezmoi.yaml) in~/.config/chezmoi/is an error, not "first one wins". When migrating formats, delete the old file in the same commit. - Config format is detected by extension. Force it with
--config-formatif you pipe a config from an unusual path. chezmoi <unknown>looks forchezmoi-<unknown>on$PATH(git-style plugin convention) before failing. Useful for adding repo-local helpers; surprising when a typo silently runs an unrelated binary.
Reference URL index
The chezmoi docs evolve fast — new template functions, new config keys,
new prefixes every few releases. When a question goes deeper than
this skill, fetch the live page rather than guessing from training
data. Prefer mcp__tavily__tavily_extract (single URL, extract_depth: "advanced", format: "markdown") — it returns the canonical
/reference/* content directly. Use Context7 only as a fallback. Do
not tavily_map or crawl — the URLs below are stable; pick the exact
one.
Example call shape:
mcp__tavily__tavily_extract(
urls: ["https://www.chezmoi.io/reference/templates/functions/jq/"],
extract_depth: "advanced",
format: "markdown",
)
Map of what lives where:
| Topic | URL |
|---|---|
| Concepts (source / target / destination / working tree) | /reference/concepts/ |
| Source-state attribute order per target type | /reference/source-state-attributes/ |
| Target types (file / dir / symlink / script / modify / create / remove) | /reference/target-types/ |
Application order on chezmoi apply | /reference/application-order/ |
| All command-line flags (global / common / developer) | /reference/command-line-flags/ |
Per-command reference (chezmoi <name>) | /reference/commands/<name>/ |
| Configuration file structure | /reference/configuration-file/ |
| Hooks (read-source-state / apply / etc.) | /reference/configuration-file/hooks/ |
| Interpreters for scripts by extension | /reference/configuration-file/interpreters/ |
Special files (.chezmoiignore etc.) | /reference/special-files/<name>/ |
Special directories (.chezmoidata/ etc.) | /reference/special-directories/<name>/ |
Template variables (.chezmoi.os etc.) | /reference/templates/variables/ |
Template directives (chezmoi:template:, etc.) | /reference/templates/directives/ |
| Built-in template functions | /reference/templates/functions/<name>/ |
init template functions (promptBoolOnce etc.) | /reference/templates/init-functions/<name>/ |
| Password-manager template functions | /reference/templates/<vendor>-functions/<name>/ (1password, bitwarden, dashlane, doppler, ejson, gopass, keepassxc, keeper, keyring, lastpass, pass, passhole, protonpass, vault, aws-secrets-manager, azure-key-vault, secret) |
| GitHub template functions | /reference/templates/github-functions/<name>/ |
| Plugins | /reference/plugins/ |
Heuristic: when the user asks "what does prefix X do" → source-state-
attributes; "what does chezmoi <cmd> do" → commands/<cmd>; "what
template var holds Y" → templates/variables; "how do I read a secret from
Z" → templates/<vendor>-functions. Fetch only the page you need with
mcp__tavily__tavily_extract — single URL, advanced depth — so the
answer is always grounded in the canonical docs as of today, not what
this skill happened to know when it was last edited.
FAQ
What does the chezmoi skill do?
Manage dotfiles with chezmoi — the source-state dotfile manager that uses filename attributes (dot, private, encrypted, runonce), Go templates, and password-manager integration to keep one set of dotfiles working across many machines. Use when the user says "set up dotfiles", "manage my dotfiles", "bootstrap a new machine", "chezmoi", "encrypt my SSH key in dotfiles", "template dotfiles per OS", "add secrets to dotfiles", "what should be in my .chezmoi.toml.tmpl", or asks how to share configs across macOS / Linux / servers. Also trigger when the user is staring at a fresh machine and wants the one-liner that gets them home, or when they're about to commit something sensitive into a dotfiles repo. Do NOT use for stow / yadm / rcm or other dotfile managers, generic git repo setup, or password-manager setup unrelated to dotfiles.
How do I install chezmoi?
Run npx skills add https://github.com/paulnsorensen/skillz-that-grillz --skill chezmoi in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/chezmoi/ (personal) or .claude/skills/chezmoi/ (project).
Which AI agents can use chezmoi?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is chezmoi safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains chezmoi?
It is published by paulnsorensen in paulnsorensen/skillz-that-grillz. It has 58 recorded installs.
More from paulnsorensen/skillz-that-grillz
skillz
paulnsorensen/skillz-that-grillz
Add, improve, audit, wedge, or autoimprove a skill. Improve and audit also accept a sub-agent definition. The result runs predictably on Claude Code, Codex, OMP, and other Agent Skills hosts. Use for /skillz , "improve this skill", "autoimprove this skill", "optimize this skill", "tighten this skill", "audit this agent", "new skill for X", "skill not triggering", "fix trigger rate", "wedge this", "turn this repeated work into a CLI", or "what in this skill should be a CLI". Do NOT use for CLAUDE.md or system-prompt edits, or for code changes that a cheese pipeline skill owns.
gh
paulnsorensen/skillz-that-grillz
Complete GitHub tasks using the gh CLI — pull request inspection, review, and merge, issues, CI checks, releases, workflow runs, code search, repo and label management. Use when the user says "merge PR", "check CI", "list issues", "review PR", "PR status", "close issue", "trigger workflow", "view release", "search repos", or invokes /gh. Use git (log, diff, status) for read-only local context only. Do NOT use for committing, staging, pushing, or PR creation — use /plate for those. Do NOT use for code-quality review — use a dedicated review skill.
prek
paulnsorensen/skillz-that-grillz
Onboard prek (Rust-powered pre-commit) into any project and suggest hooks tailored to the language/framework. Use when the user says "add prek", "set up pre-commit hooks", "configure prek", "what hooks should I use", or when you notice a project without a prek.toml that would benefit from pre-commit checks. Also trigger when the user wants to add new hooks to an existing prek.toml, audit their hook config, or migrate from pre-commit to prek. Covers Rust, Python, TypeScript/JavaScript, Go, Ruby, and Shell projects. Do NOT use for husky, lint-staged, or CI pipeline hook configuration — this skill is specifically for prek.
serena-config
paulnsorensen/skillz-that-grillz
Open the skill to read its SKILL.md instructions.
safe-settings
paulnsorensen/skillz-that-grillz
Set up GitHub's safe-settings (a Probot-based GitHub App from the github org) for declarative, org-wide repository policy as code. Use when the user says "settings as code", "safe-settings", "manage repos declaratively", "centralize repo settings", "org-level branch protection", "scaffold an admin repo", "github settings app", or invokes /safe-settings on an org that wants to manage repo settings, branch protection, rulesets, labels, teams, environments, and collaborators from one admin repo. Scaffolds the admin-repo layout (settings.yml + suborgs/ + repos/), the GitHub App installation steps, and a scheduled full-sync GitHub Actions workflow. Distinct from /gh (per-task ops) and /gh-bootstrap (one-shot gh api config of a single repo) — this skill is for org-scale settings as code that reconciles continuously.
Popular code quality & workflow skills
grill-me
mattpocock/skills
A relentless interview to sharpen a plan or design.
improve-codebase-architecture
mattpocock/skills
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
handoff
mattpocock/skills
Compact the current conversation into a handoff document for another agent to pick up.
domain-modeling
mattpocock/skills
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
codebase-design
mattpocock/skills
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
mattpocock/skills
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.