links: site | repo | project page
AI Blueprint 0.6.0 adds an explicit /ci step for setting up GitHub
verification.
I kept it separate from onboarding on purpose. /onboard and /adopt should
help a project enter the workflow without suddenly creating automation or
inventing a test suite. They now report the checks a project already has and
point to /ci when GitHub verification is wanted.
The new command builds one stack-aware Verify job from the scripts that
actually exist in the repository. If the app has linting, type checking, tests,
or a production build, those can become the gate. If it has no tests, Blueprint
does not pretend otherwise. Tests remain an explicit choice through /tests,
which can update the existing Verify job later.
Existing GitHub Actions also need to be treated as project code. /ci aligns a
current verification workflow when it can do so safely and creates
.github/workflows/verify.yml only when the repository needs one. It does not
replace unrelated automation just to force everything into a Blueprint shape.
This closes a gap in the workflow. Blueprint already had planning, feature specs, implementation checks, audits, completion gates, and release preparation. CI was something the project might have, but the workflow did not offer a clear path for creating it.
The release passed the full Blueprint validation suite, including installer
tests and packed smoke tests for Codex, Claude Code, and both adapters.
create-ai-blueprint@0.6.0 is now published on npm.