Skill installs Python projects oddly

The bundled skill has rules that tell the agent to handle Python components via runtime hooks:

https://github.com/canonical/template-sdk/blob/main/.github/skills/sdk-designer/references/sdk-patterns.md#parts-strategy-decision

This probably makes a lot of sense in some cases, but perhaps not all.

Using this repo's skill to package the [Tau coding agent](https://twotimespi.dev/), I noticed that my resulting SDK image is effectively empty, because Tau is written in Python and so installed at runtime:

https://github.com/barryprice/tau-workshop-sdk/blob/main/hooks/setup-project

This is quite a different experience to other coding agents written in compiled languages, which will land in `/var/lib/workshop/sdk` with write-protection. The Tau SDK gives the user the tool in a dotfile-dir (`/home/workshop/.tau/venv`) where it's user-writeable.

Read the referenced decision in .github/skills/sdk-designer/references/sdk-patterns.md, then inspect the linked hooks/setup-project example to understand how Python components are installed. Compare that behavior with the intended SDK image layout and write down the strategy and acceptance criteria for handling Python projects without leaving the resulting image effectively empty.

Written by the indexing model from the issue text.