Repository navigation
Adopt the analysis-workflow plugin; commit notebook outputs; zarr - #14
Conversation
- Enable the analysis-workflow plugin in .claude/settings.json and depend on its package; FilePaths.ROOT comes from its main_checkout(). - Notebook outputs are committed: drop the nbstripout filter, its install step and CI check. - Examples write AnnData zarr; *.zarr ignored. - data/<dataset>/results/ is a notebook's own outputs; data/ ignored except READMEs. - Task-output rules in .gitignore; REVIEW_GUIDE.md (generic part from the plugin). - AGENTS.md points at the plugin instead of restating conventions. The lock file is regenerated once analysis-workflow v0.1.0 is tagged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Marius1311
left a comment
There was a problem hiding this comment.
Good overall, some duplication left. Is the cruft setup in already?
|
|
||
| Humans work in notebooks; coding agents work in scripts, one task per session, git worktree and | ||
| `analysis/<topic>/.../<name>_vN/` directory. The conventions come from the | ||
| [analysis-workflow](https://github.com/quadbio/analysis-workflow) Claude Code plugin, which this |
There was a problem hiding this comment.
is it a plugin or a skill? I thought a plugin bundles several skills?
There was a problem hiding this comment.
A plugin. A plugin is the installable unit and can bundle any number of skills, plus hooks, agents and MCP servers. This one bundles one skill (also called analysis-workflow) and the guard hooks. The Python package is separate: pixi installs it, not Claude. The README now says "plugin (a skill plus guard hooks)". a8ce86b
|
|
||
| ## 🤖 Working with coding agents | ||
|
|
||
| Humans work in notebooks; coding agents work in scripts, one task per session, git worktree and |
There was a problem hiding this comment.
Looks like we're repeating this three times, here, in the skill, and in the agents.md file
There was a problem hiding this comment.
Agreed, removed from both the README and AGENTS.md. The skill owns it; AGENTS.md now only lists what the plugin owns and points there. a8ce86b
|
|
||
| --- | ||
|
|
||
| ## ☕ Daily Workflow |
There was a problem hiding this comment.
Consider how much of this is needed, given what's in the skill now.
There was a problem hiding this comment.
Cut to the install step, which is the one thing the skill can't tell a human: it only loads after installation. Dropped the package sentence (pixi.toml shows it) and the REVIEW_GUIDE.md pointer (AGENTS.md has it). a8ce86b
The README keeps only the plugin install step and points at AGENTS.md; AGENTS.md names what the plugin owns without restating it, and drops the worktree commands the skill and its hook already own. The data layout now lives once for humans, in the README; the dataset README and the sample notebook point there. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Also in a8ce86b: the Cruft: not yet. It's planned as its own PR once this one merges, so its diff (moving everything under |
Changes
.claude/settings.jsonenables it (via thequadbio/claude-pluginsmarketplace),pixi.tomldepends on its package (tag v0.1.0), andFilePaths.ROOTcomes from itsmain_checkout()instead of a local resolver.AGENTS.mdpoints at the plugin instead of restating conventions; README gains a short "Working with coding agents" section.*.zarrignored.data/<dataset>/results/is a notebook's own outputs, prefixed with the notebook's stem.data/is ignored except its READMEs and placeholders (AGENTS.md already claimed this; only extensions were ignored)..gitignore:outputs/,logs/and re-includedresults/evidence.REVIEW_GUIDE.mdfor GitHub review agents, generic part from the plugin.The lock re-solve also floats a few packages (spatialdata 0.7.3 → 0.8.0, dask, numba, ome-zarr).
Next: converting the template to cruft, so repos can pull template updates with
cruft update, as a separate PR.🤖 Generated with Claude Code