agentleFS
moeru-aimoeru-ai/auv/.agents/skills/use-pixi/SKILL.md

use-pixi

Skill207 starsChanged 38 days agoApache-2.0View on GitHub

Use when explaining, creating, or changing a Pixi workspace, including manifests, dependencies, platforms, targets, features, environments, tasks, lockfiles, editable Python packages, CI commands, or IDE interpreter setup.

  • Installs packages

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 8 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use Pixi as the command and environment boundary for this repository.

    line 8
  2. Very specific:

    Run pixi --version, pixi info, and pixi task list when available.

    line 12
  3. Very specific:

    Read the selected manifest, its workspace.platforms, targets, features, environments, tasks, and editable path dependencies.

    line 14
  4. Very specific:

    Read pixi.lock when reproducibility or CI behavior matters.

    line 15
  5. Very specific:

    Put Python itself and command-line/system packages supplied by conda-forge under Pixi dependencies.

    line 19
  6. Very specific:

    Put Python distributions resolved from PyPI under pypi-dependencies or [project.dependencies], according to the repository's existing …

    line 20
  7. Very specific:

    Use target.<platform> for dependencies limited to one supported platform.

    line 22
  8. Distinctive:

    Use a feature when a dependency set is optional and reusable.

    line 23
  9. Very specific:

    Do not use bare pip install inside a Pixi-managed repository.

    line 24
  10. Distinctive:

    Use the smallest applicable commands:

    line 37
  11. Very specific:

    When adding or removing a dependency, use --feature FEATURE for a named feature and --platform PLATFORM for a platform-specific table.

    line 52
  12. Distinctive:

    Keep the lockfile immutable in CI:

    line 56
  13. Distinctive:

    Keep all declared platforms solvable.

    line 73
  14. Distinctive:

    Do not add CUDA or another platform-only dependency to unconditional common dependencies.

    line 74
  15. Distinctive:

    Do not remove or regenerate unrelated environments.

    line 75
  16. Distinctive:

    Do not claim success until the lockfile is current and the requested task passes.

    line 76
  17. Very specific:

    Keep the default environment to Ruff and ty with no python, solvable on linux-64, osx-arm64, and win-64.

    line 83
  18. Very specific:

    Run pixi run --locked check before handoff when Python files or Pixi metadata change.

    line 86

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Use Pixi

Use Pixi as the command and environment boundary for this repository.

Inspect before editing

  1. Run pixi --version, pixi info, and pixi task list when available.
  2. Locate both pixi.toml and pyproject.toml. Pixi discovers an explicit --manifest-path first, then a local pixi.toml, then a local pyproject.toml, then parent manifests. A sibling pixi.toml therefore wins over pyproject.toml.
  3. Read the selected manifest, its workspace.platforms, targets, features, environments, tasks, and editable path dependencies.
  4. Read pixi.lock when reproducibility or CI behavior matters. Do not hand-edit the lockfile.

Choose the dependency owner

  • Put Python itself and command-line/system packages supplied by conda-forge under Pixi dependencies.
  • Put Python distributions resolved from PyPI under pypi-dependencies or [project.dependencies], according to the repository's existing ownership boundary.
  • Represent a local Python project as { path = ".", editable = true } or another relative editable path in Pixi.
  • Use target.<platform> for dependencies limited to one supported platform.
  • Use a feature when a dependency set is optional and reusable; compose features into named environments.
  • Do not use bare pip install inside a Pixi-managed repository. Use pixi add, a manifest edit, and a new lock instead.

Understand the model

  • A workspace declares channels and every platform solved into pixi.lock.
  • A target narrows configuration to a platform such as linux-64, win-64, or osx-arm64.
  • A feature groups optional dependencies, tasks, activation, and targets.
  • An environment selects features and produces an installable prefix under .pixi/envs/ by default; named or detached configuration may select another prefix.
  • A task is a reproducible command executed inside an environment.
  • The lockfile records exact conda and PyPI resolutions for every declared platform.

Make and verify changes

Use the smallest applicable commands:

pixi add PACKAGE
pixi add --pypi PACKAGE
pixi remove PACKAGE
pixi remove --pypi PACKAGE
pixi update PACKAGE
pixi install
pixi run TASK
pixi shell  # interactive use only
pixi task list
pixi lock --check

When adding or removing a dependency, use --feature FEATURE for a named feature and --platform PLATFORM for a platform-specific table. Combine the selectors only when both scopes apply.

For local, intentional manifest changes, run pixi lock, the relevant pixi run task, and pixi lock --check. Inspect the manifest diff and lockfile diff before reporting success.

Keep the lockfile immutable in CI:

pixi install --locked
pixi run --locked TASK

--locked prevents lockfile updates; it does not make environment installation or task execution read-only. Inspect task definitions before running them in any no-write workflow.

CI must not run pixi lock, because that command mutates the lockfile.

IDE boundary

Run pixi install before selecting an interpreter. In AUV, .pixi/envs/default only carries Ruff and ty; select the named environment that owns a package (under .pixi/envs/) as the VS Code python.defaultInterpreterPath folder when analyzing or debugging it. Locate named or detached environments using the prefix reported by pixi info --json, but never commit the returned absolute machine path. Documentation, automation, and terminal commands should use pixi run; Windows and POSIX interpreter paths differ.

Safeguards

  • Keep all declared platforms solvable; do not validate only the current host.
  • Do not add CUDA or another platform-only dependency to unconditional common dependencies.
  • Do not remove or regenerate unrelated environments.
  • Do not claim success until the lockfile is current and the requested task passes.

AUV workspace

  • Pixi is required only for Python or model work. Do not ask Rust, JavaScript, Protobuf, or documentation contributors to install it.
  • The root pixi.toml owns Python tooling; Rust, pnpm, and Buf keep their own toolchains. Never add them to pixi.toml.
  • Python package metadata owns runtime dependencies. Edit the package pyproject.toml, then run pixi lock.
  • Keep the default environment to Ruff and ty with no python, solvable on linux-64, osx-arm64, and win-64. Do not add a workspace-wide Python pin.
  • Give each Python package its own feature that pins python and its dependencies, and compose it into a named environment. Do not share a solve-group between packages that need different Python versions.
  • For a package on another Python version, add a directory ruff.toml that extends the root one with its own target-version, and run ty in that package's environment.
  • Run pixi run --locked check before handoff when Python files or Pixi metadata change.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.