jdx Release Notes

This release adds experimental task-scoped secrets backed by fnox, automatic updates for global tools, and an experimental opt-in install layout that names installations by what they contain. It also adds dotfile merge entries, headers auth for the http: backend, npm installs from git, and a set of prune, install and lockfile fixes.

Added

Secrets (experimental)

These need mise settings experimental=true and fnox 1.39.0 or newer. They are refused in safe mode.

• [secrets.fnox] and mise secrets ls. A project can name fnox as its secrets source in its own mise.toml. mise secrets ls lists key names and metadata but never values. -J prints JSON. Global, system and home-level [secrets] config is ignored, and mise doctor reports it. #13967

• Tasks get only the secrets they list. Add secrets = [...] to a task, or #MISE secrets=[...] in a file task header. Only that task gets the values, only while it runs, and they are redacted from its output. Dependencies, hooks, mise env, hook-env and shims get nothing. A task without grants never calls fnox. Tasks that list secrets need a trusted config, aren't artifact-cached, and ignore --raw unless the task sets raw or interactive. Hooks, watch_files, daemons and mise bootstrap can't run them, and remote or global-config tasks can't list secrets. #13974

fnox 1.39.0 lets other tools, starting with mise, get the same environment that fnox exec would give a command. They can run fnox env --json, or use the new fnox-client crate to read it straight from a running daemon's cache.

Added

• fnox env --json (#936) by @jdx. Prints a versioned JSON document describing the environment for a command. It has set (variables), files (as_file contents for the caller to write, so fnox leaves no files behind), remove (variables to unset), missing and leases. Callers apply remove, then set, then files.

  fnox env --json --keys DATABASE_URL,GCP_SA_JSON
  fnox env --json --describe   # list keys and where they can be injected, without resolving anything
  

  • --for exec (the default) covers env = true and env = "exec" secrets plus leases. --for shell covers only env = true secrets and runs no leases.
  • --keys (repeatable, comma-separated) resolves only the listed keys and the secrets they depend on. env = false secrets are never printed.
  • Errors are also printed as JSON on stdout (config, invalid_keys, resolution) and exit with status 1.
  • --describe never contacts the daemon, creates leases or prompts. With --keys, it checks the keys up front. Its output includes daemon_enabled. …

fnox 1.38.1 publishes Windows release binaries again. The v1.37.0 and v1.38.0 releases did not include them.

Fixed

• Windows x64 and ARM64 release archives (#934) by @jdx. The Windows release builds for v1.37.0 and v1.38.0 failed while building the bundled OpenSSL, so neither release shipped Windows binaries. The release workflow now builds OpenSSL with the runner's native Perl, and Windows users can download prebuilt binaries again from this release onward. fnox itself has no code changes since v1.38.0.

Full Changelog: https://github.com/jdx/fnox/compare/v1.38.0...v1.38.1

💚 Sponsor fnox

fnox is built and maintained by @jdx, an open source developer at entire.io, the title sponsor of his open source work.

If fnox handles secrets or config for you or your team, please consider becoming an individual or company sponsor. Your support funds ongoing development and helps keep fnox secure, free, and independent.

This release adds dotfile groups for Stow-style dotfiles repositories, templated [bootstrap.compose] values, and custom labels for task confirm prompts. Secret values no longer get copied into __MISE_DIFF/__MISE_SESSION or written to the env cache. It also fixes dotenv ${VAR} expansion, mise x tool@latest reporting other tools as missing, and Ruby source builds ignoring depends tools.

Added

• Dotfile groups. You can now declare a directory tree as a group with [dotfile_groups.<name>], using one directory per app or machine role like GNU Stow. mise walks the group's root and deploys each file to the same path under target (default ~), so you don't have to list files one by one. A machine picks which groups to apply with [bootstrap] dotfile_groups. If that list isn't set, every group applies. #13945

  [dotfile_groups.home]
  root = "home"            # relative to dotfiles.root
  dot_prefix = true        # dot-config/ deploys as .config/
  exclude = ["README.md"]
  
  [dotfile_groups.home.entries]
  "~/.config/kitty" = { mode = "symlink" }                            # link this directory whole
  "~/.gitconfig" = { source = "git/config.tmpl", mode = "template" }
  
  [bootstrap]
  dotfile_groups = ["home", "zsh"]
  ``` …
  

hk 2.5.0 makes stashing safe when hk is killed and when several hk processes run at once, loads configs much faster, and reports diagnostics from many more builtins. It also fixes a long list of cases where a check passed without checking anything, along with several Windows problems.

Highlights

• Stashing holds up under interruption and concurrency. SIGTERM, SIGHUP and crashes no longer leave your unstaged changes stuck in git stash. Hooks running in linked worktrees now take turns stashing. Several edge cases that dropped or mixed up changes are fixed.
• Configs load faster. A cold config load drops from about 1.4 s to about 0.07 s, and hk validate is about 14x faster. Several per-step and stashing costs are lower in large repos.
• Fewer checks pass when they should fail. hk validate now catches more config mistakes. Several builtins and hk util checks no longer skip binary files, case conflicts in directory names, or executable bits recorded in git.

Added …

This release adds an experimental spinel: backend for compiling Ruby CLIs to native binaries. The mise.toml schema now checks each backend's own tool options, and daemon presets work on Windows. It also includes fixes for shell activation with untrusted configs, task timeouts and Windows quoting.

Added

• Experimental spinel: backend. It builds a Ruby command-line tool from a GitHub repository into a native executable using Spinel, Matz's Ruby AOT compiler. Versions come from git tags through git ls-remote, so listing them doesn't call the GitHub API. Available options: entrypoint, bin, tag_prefix, source_ref and spinel. You need the spinel compiler on PATH (mise doesn't install it yet) and mise settings experimental=true. It works on macOS and Linux only, and it may be removed later if it becomes a maintenance burden. Based on nateberkopec/mise-backend-spinel. #13922

  [tools."spinel:tobi/try"]
  version = "1.10.1"
  entrypoint = "try.rb"
  bin = "try"
  tag_prefix = "v"
  ``` …
  

This release is mostly bug fixes. mise run now stops tasks properly when --timeout expires or you press Ctrl-C. Task daemons and native shims work better on Windows. core:rust now follows mise.lock and picks up new stable and beta toolchains. Lockfile, GitHub asset selection, Homebrew cask and plugin update problems are also fixed.

Fixed

Tasks

• mise run --timeout now stops the tasks it was running. Before, mise printed the timeout error and exited, but the task processes could keep running in the background. Now the whole-run timeout (--timeout or the task.timeout setting) stops tasks the same way a per-task timeout does. On Unix, mise sends SIGTERM and then SIGKILL after 5 seconds. On Windows, it runs taskkill /F /T. Tasks with raw = true are not stopped by the whole-run timeout. #13876 by @Marukome0743
• A single Ctrl-C lets tasks shut down cleanly. Before, one Ctrl-C could make mise exit right away while tasks were still cleaning up. This happened in three cases: a task that runs mise run itself got SIGINT twice; a tool like docker compose up treated the duplicate SIGINT as a force-quit; and a task that exits non-zero on SIGINT caused mise to send SIGTERM to its sibling tasks. Now mise waits for tasks to finish and then exits with status 130. A second Ctrl-C still force-quits. #13904 …

This release tightens supply-chain verification. Keyless cosign bundles must now match a pinned signer identity, and GitHub attestation workflow checks no longer accept partial matches. It also closes a .tool-versions trust gap that could leak GITHUB_TOKEN, adds a per-cask appdir option for Homebrew casks, and fixes problems with locked SLSA installs, musl hosts and mise backends switch.

Security

• Inline tool options in .tool-versions now require trust. This is the .tool-versions version of the mise.toml fix in 2026.9.18. An untrusted project could ship a github: entry with inline options, such as [api_url=...], pointing at another host. Commands such as mise ls, env, current, outdated and latest would then send your GITHUB_TOKEN to that host without asking for trust. Any entry whose tool name contains [ now requires trust, the same as Tera templates. Plain lines like node 20.0.0 still load without trust, and a [ inside a comment is ignored. Run mise trust for projects you rely on. MISE_SAFE=1 still skips trust checks. (GHSA-wcqh-j26q-g44x) #13869 …

mise.toml can now include a shared config file from a git repository or OCI registry, and task_config.includes accepts OCI artifacts. The release also closes a trust bypass that could send GITHUB_TOKEN to an attacker-controlled host, adds gem registry sources, and fixes several daemon and dotfiles problems.

Security

• Inline tool options now require trust. Before this change, a mise.toml in an untrusted directory could hide options in a tool key, for example a github: tool key with [api_url=...] pointing at another host. mise loaded the file without trust because the value was a plain version string. Commands such as mise ls, env, current, outdated, upgrade --dry-run and latest then sent GITHUB_TOKEN to that api_url. Now any tool key that contains [ requires trust, the same as { ... } option tables already did. Plain keys like node or "cargo:eza" still load without trust. If you use inline options in a project you haven't trusted yet, run mise trust. MISE_SAFE=1 still skips trust checks entirely. #13849

Added

• Include shared config from git or OCI. Organizations can keep tool versions, env and hooks in one place and pull them into every repo: #13843

  include = [
    "git::<repo-url>//mise.toml?ref=main",
    "oci::ghcr.io/myorg/platform-config@sha256:0f1e2d3c...",
  ]
  
  [tools] …
  

mise self-update and the mise.run installer now pick the newest stable release that is at least 24 hours old. Updates also check the release's signed packslip before replacing the binary. This release also adds a machine-local global miserc, an opt-in way for command-not-found to install registry tools, and a postinstall mode that runs on every install. It fixes several Homebrew formula builds and closes a trust gap in paranoid mode.

Changed

• Self-update and installs wait for a minimum release age. When no version is pinned, mise self-update, automatic updates, update notifications, and the mise.run installer now choose the newest stable release published at least 24 hours ago. Explicit versions skip the delay. An unpinned update never downgrades a newer installation, even with --force. The age is taken from, in order: --minimum-release-age, then self_update.minimum_release_age, then the global minimum_release_age setting, then 24h. Use 0s to get releases right away. #13782

  [settings]
  self_update.minimum_release_age = "7d"
  

  mise self-update --minimum-release-age 0s
  curl -fsSL https://mise.run | MISE_SELF_UPDATE_MINIMUM_RELEASE_AGE=7d sh
  ``` …
  

fnox 1.36.0 makes the per-prompt shell hook much faster, batches fnox reencrypt by provider, fixes the YubiKey provider, which timed out on every challenge, and ships a fnox agent skill with each release.

Added

• fnox agent skill (#910) by @jdx. A new fnox skill teaches coding agents how to set up providers and profiles, inject secrets into commands, and diagnose authentication or resolution failures without printing secret values. The release Packslip now includes the skill, so you can install it with mise (you need a mise version with Packslip skill support):

  mise use packslip:github.com/jdx/fnox
  mise skills sync --dir .agents/skills
  

  The generated skill links only work on your machine, so keep them out of version control.

Fixed

• fnox reencrypt batches encryption by provider (#880) by @davdroman. reencrypt used to encrypt each secret on its own, even when the provider supports batching. For age secrets protected by Touch ID, this meant one unlock per value on an uncached exec. fnox now makes one batch encryption call per provider. You can also use this to merge secrets that were encrypted separately, such as new values added with fnox set:

  fnox reencrypt --provider age
  

  Secrets you don't select stay unchanged, source files and profiles are kept as they were, and if a provider batch fails, fnox saves none of that batch's changes.…

In hk fix, many more builtins now apply their check_diff patch instead of running the tool a second time as the fixer. hk now applies those patches itself instead of calling git apply, and read-only check_diff commands can run in parallel with other steps that read the same files.

Added

• hk util format-diff (#1514, @jdx). Builds an applicable patch from any formatter that reads stdin and writes the formatted file to stdout. It runs the formatter once per file, in parallel, and replaces {} with the file's path. It exits 1 with a patch when a file would change, and 0 when nothing would. If the formatter fails for any file or prints nothing for a non-empty file, it prints no patch, so hk runs fix and that reports the error. Use --no-stdin for formatters that read the file from its path.

  check_diff = "hk util format-diff {{files}} -- stylua --stdin-filepath {} -"
  

• hk util sarif-diff (#1515, @jdx). Runs a tool that reports fixes in a SARIF log and turns them into a patch. It prints a patch only when every result has a usable fix. If any result has no fix, or the tool fails, it prints no patch and hk runs fix. The findings exit code defaults to 1 and can be changed with --findings-exit-code.

  check_diff = "hk util sarif-diff -- pinact run --check --format sarif {{files}}"
  ``` …
  

Aqua tools can now choose glibc or musl builds one tool at a time, and monorepo roots can get short task path aliases. Packslip tools keep installing after their GitHub or GitLab repository is renamed, because mise now pins them by repository ID, recorded in a new lockfile revision 3. SLSA provenance checks now require the expected signer identity. This release also fixes regressions in mise run --no-timings, cargo +nightly and the outdated/upgrade version comparison, and speeds up shims and config loading.

Added

• Per-tool libc for aqua tools. On glibc Linux, mise prefers a release's gnu build even when the aqua registry names the musl one. That breaks tools whose musl build is the fully static one, such as aqua:domcyrus/rustnet. You can now pick the build for a single tool instead of changing the global libc setting. #13701

  [tools]
  "aqua:domcyrus/rustnet" = { version = "latest", libc = "musl" }
  ``` …
  

hk now starts faster, spends less of its own time per file, and spreads --jobs more evenly across batched steps. This release also fixes several cases where stashing could lose files, fail before any step ran, or leave untracked files visible to steps.

Added

• apply_check_diff step setting (#1495, @jdx). Defaults to true. If you set it to false, hk fix runs the step's fix command instead of applying the check_diff output, and hk check still shows the diff. In staging hooks such as pre-commit, hk still runs check_diff first and passes fix only the files the diff names. The trailing_whitespace and newlines builtins now use it. Together with a parallel file scan in hk util trailing-whitespace and hk util end-of-file-fixer, this roughly halves the time of those two steps on a whole-repo fix.

  ["my-fixer"] {
    check_diff = "my-fixer --diff {{files}}"
    fix = "my-fixer --write {{files}}"
    apply_check_diff = false
  }
  

Changed …

mise oci build can now package tools installed by vfox plugins, and vfox plugins can repair an existing install when its tool options change. Shell prompts, cd, and settings loading are faster. Dotfiles include/exclude patterns now follow .gitignore rules for * and a leading /, which fixes a case where rollback could delete a live file.

Added

• vfox tools in OCI images (experimental). mise oci build used to reject every tool installed by a vfox plugin. It now builds those tools into the image, with one layer per tool plus one layer per plugin at /mise/plugins/<name>/, so mise inside the image can resolve the tool without cloning the plugin. The plugin's env hook runs on the build host. Install-dir paths are rewritten to their in-image location, and mise warns when a value points into the host's home directory. Changing a plugin invalidates the reused layers of its tools on mise oci push. asdf plugins are still rejected. #13670 …

Staging hooks such as pre-commit now stage each fixer's files as soon as that fixer finishes, and hk fix --all makes better use of --jobs. This release also adds a batch_min_files step setting and fixes the jq/yq builtins, stashing before the first commit, and an intermittent git crash during concurrent staging.

Added

• batch_min_files for batched steps (#1475, @jdx). Steps with batch = true can set the minimum number of files hk passes to one process. The default is still 4. Raising it means fewer processes, which helps tools with slow startup, such as prettier and eslint, on whole-tree runs. In hk's benchmark, batch_min_files = 200 on prettier and eslint cut hk check --all from 1.56 s to 1.33 s. It made a small pre-commit a little slower, so the builtins' defaults are unchanged. OS command-length limits can still split a batch below this number.

  steps {
    ["prettier"] = (Builtins.prettier) {
      batch_min_files = 200
    }
  }
  

Changed …

Fix steps now run their fixer directly instead of checking first, and batch steps start fewer processes on small commits. Together these make hk run pre-commit about a third faster on a typical fixture. This release also adds JUnit XML output, hook arguments in conditions, and a hk migrate pre-commit that produces a working config right away. It fixes a staging race between concurrent fixers and a stale config cache when Pkl reads environment variables.

Added

• --junit-xml <PATH> for hk check, hk fix and hk run <hook> (#1432, @sanga). Writes one JUnit <testcase> per step, with its status, duration and captured output, so CI systems such as Jenkins can show hk results. It works alongside --sarif, whatever output format you choose. If a run fails before any step is planned, the report still contains a failing test case, so it doesn't look like a passing build.

• Git hook arguments in condition and step_condition (#1445, @jdx). Conditions can now use the hook variables that command templates already had: prev_head, new_head and is_branch_checkout (post-checkout), commit_msg_file (commit-msg, prepare-commit-msg), source and sha (prepare-commit-msg), hook_stdin (pre-push, post-rewrite), and hook_args (all hooks). Conditions are expr-lang expressions, so write is_branch_checkout, not {{ is_branch_checkout }}.

  hooks { …
  

A folder inside any conf.d directory now loads as its own config fragment and serves as the config root for the files in it, which gives [bootstrap].config_roots users a direct migration path. [dotfiles] gains two GNU Stow-style options: relative symlinks and dot-<name> sources. Release metadata for any public github.com repo now comes from mise-versions, and the registry can require GitHub attestations for specific tools.

Added

• conf.d folder fragments. A folder in a global, system, or project conf.d directory now loads as a fragment. Relative paths, {{ config_root }}, and task working directories resolve inside that folder, so a bundle can keep its files next to its config. Each folder can hold mise.toml, mise.local.toml, mise.<env>.toml, and mise.<env>.local.toml. Folders are not searched recursively, and folders whose names start with . are skipped. A folder can be a symlink. Folder fragments load after the single-file fragments in the same conf.d (in folder-name order) and before config.toml. mise use/mise set never write to them. #13603

  ~/.config/mise/conf.d/
  ├── git.toml          # single-file fragment, unchanged
  └── git-tools/        # folder fragment
      ├── mise.toml
      └── gitconfig
  

  # ~/.config/mise/conf.d/git-tools/mise.toml
  [dotfiles]
  "~/.gitconfig" = "gitconfig"   # resolves to conf.d/git-tools/gitconfig
  ``` …
  

Fix steps now run their fixer directly instead of checking first, and batch steps start fewer processes on small commits. Together these make hk run pre-commit about a third faster on a typical fixture. This release also adds JUnit XML output, hook arguments in conditions, and a hk migrate pre-commit that produces a working config right away. It fixes a staging race between concurrent fixers and a stale config cache when Pkl reads environment variables.

Added

• --junit-xml <PATH> for hk check, hk fix and hk run <hook> (#1432, @sanga). Writes one JUnit <testcase> per step, with its status, duration and captured output, so CI systems such as Jenkins can show hk results. It works alongside --sarif, whatever output format you choose. If a run fails before any step is planned, the report still contains a failing test case, so it doesn't look like a passing build.

• Git hook arguments in condition and step_condition (#1445, @jdx). Conditions can now use the hook variables that command templates already had: prev_head, new_head and is_branch_checkout (post-checkout), commit_msg_file (commit-msg, prepare-commit-msg), source and sha (prepare-commit-msg), hook_stdin (pre-push, post-rewrite), and hook_args (all hooks). Conditions are expr-lang expressions, so write is_branch_checkout, not {{ is_branch_checkout }}.

  hooks { …
  

mise run can now export OpenTelemetry traces and logs (experimental), and experimental daemon providers let several projects and worktrees share one PostgreSQL, CockroachDB, or NATS server, each with its own database or account. Lockfiles no longer switch backends on their own when the registry moves a tool: the new mise backends switch command does it when you ask. [dotfiles] and [bootstrap.files] can now remove files and manage permissions, and mise bootstrap unapply removes what a module set up. The experimental pkgx: backend has been removed.

Highlights

• Observability and shared services (experimental): task runs export OTLP traces and, if you opt in, task output as logs. Global [daemon_providers] run long-lived servers, and projects attach to them with an isolated database or NATS account per checkout.
• Safer lockfiles: locked tools stay on their locked backend, mise lock --bump checks remote versions and fails when it can't, lockfiles no longer record versions that were never confirmed, and tool stubs lock into the project's mise.lock.
• Declarative cleanup: mode = "absent", remove_empty templates, permissions-only entries, removal of empty directories mise created, and mise bootstrap unapply let a config describe what should not be on a machine.

Added

Tasks …