Agent skills for Bayesian modeling with PyMC, PyTensor and ArviZ. Each skill is a folder of instructions and focused references in the Agent Skills format.
| Skill | Use it for |
|---|---|
| pymc-modeling | Model specification, inference, predictions and specialized model families |
| prior-elicitation | Prior selection, elicitation, predictive checks and shrinkage |
| arviz-diagnostics | MCMC diagnostics, predictive checks, LOO and model comparison |
| pytensor-workflows | Symbolic graphs, shapes, compilation, gradients and custom Ops |
From the repository root, use the Skills CLI:
npx skills add .Choose the skills and agents you want.
Or copy individual directories from skills/ into your agent's supported skills
location. Keep the whole directory, including references, any scripts and LICENSE.
No custom installer or running service is required.
Use the following commands only after the matching skills and marketplace
manifest have been published to fonnesbeck/pymc-agent-skills. A local checkout,
installer discovery or manifest validation does not establish remote availability
or successful plugin installation.
For the Skills CLI:
npx skills add fonnesbeck/pymc-agent-skillsFor a Claude Code managed bundle:
/plugin marketplace add fonnesbeck/pymc-agent-skills
/plugin install pymc-agent-skills@pymc-agent-skills
Choose either the plugin or the Skills CLI/manual installation, not both, to avoid duplicate skills.
Ask your agent to perform a relevant task, or invoke the skill by name using your agent's skill mechanism. For example:
- “Build a hierarchical model for these grouped observations.”
- “Check whether these priors imply plausible outcomes.”
- “Diagnose this posterior and assess whether LOO is reliable.”
- “Find the shape or gradient error in this PyTensor graph.”
The guidance targets PyMC 6+, PyTensor 3+ and ArviZ's DataTree API. Code examples use the consuming project's compatible Python environment. Optional packages such as PreliZ, pymc-extras, BART and alternative backends are needed only for the workflows that use them. Consult version-matched documentation; do not replace a working environment just to install these instructions.
Adapted from pymc-labs/pymc-modeling. Licensed under MIT. Each skill includes the same license for standalone distribution, including the original PyMC Labs copyright notice.