devenv-project
Skill9 starsChanged 55 days agoMITView on GitHub
Create, explain, review, and debug devenv project environments, including devenv.nix, devenv.yaml, lockfiles, languages, tasks, services, and shell activation.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- home-managerSkill
- microvm-nixSkill
- nix-darwinSkill
- nix-languageSkill
- nixos-coding-agentsSkill
- nixos-operationsSkill
- nixos-wikiSkill
- nixpkgs-developmentSkill
- nix-workflowSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 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.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Read the project's devenv.nix, devenv.yaml, devenv.lock, applicable local configuration, and repository instructions before editing.
- Very specific: line 13
Check devenv version, the locked devenv modules and nixpkgs revisions, and require_version when present.
- Distinctive: line 20
Read only the references needed for the task:
- Very specific: line 30
Keep project dependencies in project configuration.
- Distinctive: line 33
Do not delete it to hide evaluation failures.
- Very specific: line 54
Never search /nix/store (for example find /nix/store/*foo-* -name libfoo.so) and never copy a literal store path.
- Very specific: line 58
Do not quote attribute names that are valid identifiers ([A-Za-z_][A-Za-z0-9_'-]*, dashes included).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
devenv project environments
Read the project's devenv.nix, devenv.yaml, devenv.lock, applicable local
configuration, and repository instructions before editing. Preserve existing
conventions and overrides. For a new project, use devenv init when available;
inspect its output before adding the smallest configuration needed.
Check devenv version, the locked devenv modules and nixpkgs revisions, and
require_version when present. These versions can differ independently. The
bundled references describe the release in sources.json, not
necessarily the user's project. Verify version-sensitive options against its
actual modules. devenv info evaluates configuration; devenv search NAME uses
the project's package input. Evaluation can fetch or build dependencies.
Read only the references needed for the task:
- Configuration: files, packages, inputs, pins, imports, profiles, and YAML version requirements.
- Workflows: native activation versus direnv, scripts, task dependencies, processes, tests, and git hooks.
- Options: selected option types, defaults, declarations, and Python/JavaScript/Rust/Go/PostgreSQL examples. For other options, consult the target revision's modules or documentation; this is not a full catalogue.
Keep project dependencies in project configuration. Host installation follows
local platform policy. Update locks deliberately (devenv update NAME for one
input), inspect the diff, and preserve the lock in version control as appropriate.
Do not delete it to hide evaluation failures. Package versions come from project
inputs; changing the host toolchain does not change those pins.
Distinguish native devenv hook directory trust (devenv allow) from direnv's
.envrc approval (direnv allow). Inspect the active integration and shell setup
before changing either. Do not silently grant trust, edit startup files, execute
unknown hooks, or start services merely to inspect a project. Respect task
authorization already granted; do not repeatedly request the same consent.
Choose validation according to side effects: devenv info evaluates;
devenv shell -- COMMAND also runs shell initialization; tasks execute their
dependencies; devenv up starts processes; devenv test runs enterTest and can
start and stop configured processes. Inspect hooks/tasks first and report what
was actually run. Keep secrets out of committed configuration and the Nix store.
Guidance was informed by upstream's setup skill.
References contain modified upstream excerpts under Apache-2.0.
This portable skill does not replace a machine-specific devenv policy skill.
Nix style
- Never search
/nix/store(for examplefind /nix/store/*foo-* -name libfoo.so) and never copy a literal store path. Find the providing package withnix-locate, and refer to it through Nix:${pkgs.foo}/liborlib.makeLibraryPath [ pkgs.foo ]. - Do not quote attribute names that are valid identifiers
(
[A-Za-z_][A-Za-z0-9_'-]*, dashes included). Writepkgs.foo-barandpackages.x86_64-linux, notpkgs."foo-bar". Quote only other names (".config/foo","2.0"), the keywordsassert else if in inherit let or rec then with, and interpolations ("${name}"). Upstream examples in references sometimes quote needlessly; do not copy that style.
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.

