Skip to content
Agentmart
Submit a tool

chezmoi

Manage chezmoi source names, templates, secrets, diffs, and deployment in fmind/dot.

Skill by Médéric HURIER (Fmind)

IndexedMIT★ 12
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/fmind-chezmoi

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Manage chezmoi source names, templates, secrets, diffs, and deployment in fmind/dot.

Use cases: security

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/chezmoi

What's inside

Skill name
chezmoi
Description
Manage chezmoi source names, templates, secrets, diffs, and deployment in fmind/dot.
License
MIT
Metadata
kind: task · author: Médéric HURIER (Fmind) · source: github.com/fmind/dot/tree/main/.agents/skills/chezmoi · created: 2026-07-12 · updated: 2026-10-24
Instructions
47 lines in SKILL.md · view on GitHub

Before installing

  • Instructions read files in your home folder (~/.ssh, ~/.aws…)

Files (2)

  • SKILL.md
  • source-names.md · references/

SKILL.md (first lines)

# Chezmoi Source Standard

The source tree (`~/.local/share/chezmoi`) is the only thing to edit; `chezmoi apply` renders it into `$HOME`, so a change made under `~/.config` or `~/.gemini` is overwritten on the next apply. File names encode target path, mode, encryption, and rendering; [mise](../../../skills/mise/SKILL.md) wraps the commands as tasks.

## Naming

Read [source names](references/source-names.md) when adding or renaming a managed target; attributes depend on the target type.

## Workflow

1. **Edit the source**, never the deployed copy; `chezmoi cd` opens a shell in the source root.
1. **Manage an existing file**: `chezmoi add <target>` infers the attributes; `--template` templatizes; set a secret to `0600` before `--encrypt` imports it as `encrypted_private_<name>.age` (with `dot_` before a leading-dot name):

   ```bash
   chmod 600 ~/.config/<tool>/secret
   chezmoi add --encrypt ~/.config/<tool>/secret   # import and encrypt into the source
   chezmoi edit ~/.config/<tool>/secret            # edit the plaintext, re-encrypt on save
   ```

1. **Preview**: `mise run diff` (`chezmoi diff --force --no-pager`; restrict to affected targets). Diffs exclude encrypted targets so plaintext secrets never reach transcripts; `chezmoi status` still lists their drift.
1. **Validate rendering**: `mise run check:chezmoi` uses temporary configuration and destination with a dry run excluding encrypted files. It checks the current platform; exercise changed Linux/macOS branches separately and report unexercised branches or encrypted targets.
1. **Check repeatability**: for Bash/profile modifier changes, add existing-target and second-pass cases under `dot/tests/`, using the isolated rendering approach in `test_shell_modifiers.py`. Their blocks go through `.chezmoitemplates/managed-block.tmpl`, which rewrites the block between its begin and end markers in place, so a body change needs no migration code. JSON/TOML modifier changes need the same cases in `test_harness_config.py`. Inspect affected `run_*` hooks: ordinary hooks run on each apply; once/onchange hooks can rerun when their content changes.
1. **Apply within scope**: `mise run apply` (`chezmoi apply --force`, so a diverged target never blocks scripts on a prompt; `--force` does not expand authorized targets or side effects). Themes (one fmind/theme commit) and fonts (release URLs) are checksum-pinned externals that refetch when their URL changes; a checksum mismatch fails apply instead of deploying changed content, and `mise run upgrade` advances the theme pin only while every copied block in the `COPIED` map of `dot/dot_tasks/theme_pin.py` matches. The Google Sans, Google Sans Code, and Nerd Fonts releases have no upgrade task: when a newer release exists, bump each URL and its checksum together.
1. **Pull target edits back**: `chezmoi re-add` folds manual changes to a managed file (a regenerated lockfile, for example) into the source.
1. **Diagnose**: `mise run doctor` (`dot doctor`, `chezmoi doctor`, and `mise doctor`); `chezmoi managed` and `chezmoi unmanaged` list coverage; `dot orphan` lists former managed targets ([orphans](../../../skills/dot-cli/references/orphans.md)); the [installed-link recovery guide](../dot-skills/references/installed-links.md) covers retired skill links, and approved cleanup moves them to recoverable backups. Use command help and [dot-cli](../../../skills/dot-cli/SKILL.md) for cleanup flags.

## Gotchas

- **Attribute order is fixed**: `encrypted_` before `private_` before `dot_`; a wrong order yields a literally named file instead of the effect.
- **Keep the modify-template convention**: chezmoi supports `modify_*.tmpl` scripts. This repository instead uses `# chezmoi:modify-template` and `.chezmoi.stdin` for every modifier (shell blocks via `managed-block.tmpl`, JSON/TOML via `config-merge.tmpl`); preserve that convention unless intentionally changing the execution model.
- **Escape other tools' template delimiters**: emit another tool's `{{ ... }}` as ``{{`{{ .Destination }}`}}`` (backticks inside an action); `.chezmoi.toml.tmpl` needs this too.
- **Templates fail closed**: one template error aborts the whole apply; debug with `chezmoi execute-template < file` or `chezmoi apply --dry-run` before committing.
- **Commit only encrypted `*.age` secrets**: keep plaintext out of the repository, previews, and logs (inspect secret targets by status/metadata, never a plaintext diff); rotate a leaked secret (see [code-security](../../../skills/code-security/references/code-review/GUIDE.md)). Native login seeds use `create_encrypted_private_*` so account switches survive apply; scoped keys stay under `~/.config/dot/secrets/`, never in global shell exports. Follow [secret setup](../../../README.md#secret-management) and [credential precedence](../../../skills/dot-cli/references/authentication.md).
- **`.chezmoiignore`** (templated doublestar patterns, not gitignore syntax) keeps repo-only files (`dot/`, `skills/`, `AGENTS.md`) out of apply and skips key-dependent files without the age key; chezmoi already skips dot-prefixed sources such as `.github/`.
- **Removing a source never deletes its target**: create `.chezmoiremove` with the retired targets (glob patterns) so apply deletes them on every machine; delete each entry, and the file once empty, when no machine can still hold the target.
- **Ignore patterns match target paths**: a leading `!` excludes a match from ignoring and takes priority over every other pattern, whatever the order.
- **`.chezmoi.toml.tmpl`** seeds `~/.config/chezmoi/chezmoi.toml` on `chezmoi init` (`promptStringOnce` per-host data); `[add] secrets = "error"` makes `chezmoi add` refuse unencrypted secrets, and `[edit] apply = true` applies after `chezmoi edit` (`--watch` on save).

## Documentation

- [chezmoi reference](https://www.chezmoi.io/reference/) · [source-state attributes](https://www.chezmoi.io/reference/source-state-attributes/)
- [templating](https://www.chezmoi.io/user-guide/templating/) · [age encryption](https://www.chezmoi.io/user-guide/encryption/age/)
- Releases: [chezmoi](https://github.com/twpayne/chezmoi/releases)
- Companion skills: [mise](../../../skills/mise/SKILL.md) (pins chezmoi, wraps apply, diff, doctor), [dprint](../../../skills/dprint/SKILL.md) (formats source configurations).
- Also: [code-security](../../../skills/code-security/references/code-review/GUIDE.md) (leak scanning around `*.age` files), [dot-cli](../../../skills/dot-cli/SKILL.md) (workstation and archive commands).

From fmind/dot, MIT.

More from fmind/dot

Signals

Stars
12
Forks
1
Open issues
0
Last commit (this folder)
Oct 24, 2026 (4 days ago)
Last release
Oct 24, 2026 (4 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
Website
fmind.dev
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from Médéric HURIER (Fmind)

Cite this page

chezmoi skill. Agent Mart. https://agentmart.dev/skills/fmind-chezmoi (last updated 2026-10-24)