Onboard the user's existing Python project into Dojo. Reads their code, asks targeted questions about the data + evaluation, scaffolds `.dojo/`, writes `PROGRAM.md` + `SETUP.md` from the dialogue, and drives `dojo domain setup` until the AI-generated `load_data.py` and `evaluate.py` connectors verify cleanly against the user's data. Use when the user wants to set up Dojo on a project that already has data loaders or an evaluation metric — i.e. they're not just trying the framework on a canned sklearn dataset. Trigger phrases: "onboard my project to Dojo", "set up Dojo here", "run /dojo-onboard", "I want to use Dojo on this codebase".
End-to-end workflow that turns a one-line user prompt about a desired change into a released version — chains create-issue, solve-issue (plan), pause for plan review, solve-issue (implement) + PR, pause for PR merge, then drives the post-merge release (verify issue closure, tag, publish). Use whenever the user describes something they want shipped and gestures at the whole flow — phrases like "ship this", "/ship-it", "drive this end to end", "open an issue and work it", "let's just do the full loop", or anything that implies create-issue → plan → review → PR → release rather than just one step.
Drive a GitHub issue end-to-end: research the codebase, write a plan.md, pause for user review, then implement on a feature branch and open a PR (with version bump per docs/RELEASING.md if present). Use whenever the user says "solve issue #N", "work on issue #N", "implement issue #N", "let's tackle that ticket", or names a GitHub issue URL and wants progress on it. The skill runs in two phases — plan (stops for review) and implement — and detects which phase to run by checking for an existing plan.md for that issue.
Create a GitHub issue from a short user prompt for the Dojo repo. Use whenever the user describes a bug, feature, refactor, or improvement they want tracked — phrases like "open an issue for…", "file a ticket for…", "we should fix…", "let's track…", "create an issue", or any change request that doesn't need to be implemented right now. The skill does a brief code search to enrich the issue with file/line references, drafts a clear title + body, and creates the issue via `gh`. Optimised for speed — deep research belongs in solve-issue, not here.