OpenParable
Skip to content
← Back to skills

Python Environment

Development
python-environment

Python Environment is an Agent Skill for development published in the pymc-labs/CausalPy repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Detect, configure, and use a conda-compatible tool. Use before tasks that need the project environment, such as importing project code, running tests, building docs, or invoking repo tooling.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/pymc-labs/CausalPy/blob/HEAD/.agents/skills/python-environment/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/python-environment, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Python Environment

Set up and run commands inside the CausalPy conda environment.

Decide whether the env is required

Use the CausalPy env when the command:

  • imports project code (for example import causalpy or project modules)
  • runs tests
  • builds docs
  • invokes repo tooling such as make, prek, or notebook execution

For simple inspection helpers that only read local text/JSON or use the Python standard library, any Python on PATH is acceptable.

Reuse before creating

Do the least work that will get the task done:

  1. Reuse an existing CausalPy env if one is already available.
  2. If run -n CausalPy cannot resolve the env, check whether it exists under a different prefix and use run -p.
  3. Only create the env if no suitable existing env is available.
  4. Only update the env or rerun make setup when dependencies changed, the editable install is stale, or the current checkout has not been installed into that env yet.

Detect the conda tool

Use whichever of mamba, micromamba, or conda is available (checked in that order):

# Check for mamba, micromamba, or conda (in preference order) on $PATH
CONDA_EXE=$(for c in mamba micromamba conda; do command -v "$c" &>/dev/null && echo "$c" && break; done)

If CONDA_EXE is empty, no conda-compatible tool was found. Propose installing micromamba to the user:

"${SHELL}" <(curl -L micro.mamba.pm/install.sh)

After installation, set CONDA_EXE=micromamba.

Create the environment only if needed

If no suitable existing env can be reused, create it:

$CONDA_EXE env create -f environment.yml

Install the package only when needed

Run make setup after creating or updating the env. Also rerun it when using a different git worktree if that env has not been installed against the current checkout yet.

$CONDA_EXE run -n CausalPy make setup

Run commands

Never use $CONDA_EXE activate, instead use $CONDA_EXE run -n CausalPy <command>.

$CONDA_EXE run -n CausalPy <command>

For example: $CONDA_EXE run -n CausalPy pytest, $CONDA_EXE run -n CausalPy prek run --all-files.

Update an existing environment

$CONDA_EXE env update --file environment.yml --prune

Troubleshooting

Named env cannot be resolved

If $CONDA_EXE run -n CausalPy ... fails with errors such as The given prefix does not exist:

$CONDA_EXE env list
$CONDA_EXE run -p "/full/path/to/CausalPy" <command>

Keep using run -p with that full prefix for the rest of the session.

Git worktrees and remote machines

Git worktrees do not require a fresh env per agent session. Prefer reusing an existing env to save time. The main caveat is that this repo uses editable installs, so one shared env can point at whichever checkout most recently ran make setup.

  • For ordinary local work on one checkout, reuse the existing env.
  • For long-lived parallel worktrees, one env per worktree is the safest option, but do not create one unless needed.
  • On a fresh remote machine or ephemeral container, create the env once. On a persistent remote machine with an existing env, reuse it.

If you hit issues with an outdated tool, update it:

  • mamba / micromamba: $CONDA_EXE self-update
  • conda: conda update -n base conda

As of 2026-02-13, current versions are conda 26.1.0, mamba/micromamba 2.5.0.