Spec-driven BlueStep development with AI agents — scaffolder and project conventions for Claude Code
Agent Claude Code
Implements exactly ONE already-approved task from a feature spec in this repo (the bspecs scaffolder), in an isolated context. Invoked by /spec-execute to keep scoped source reads out of the main session. Reads the task's spec files and only the files the task references, makes the change, verifies, and returns a…
Agent
Reviews BlueStep component code and returns a structured report grouped Critical / Warnings / Suggestions — try/catch coverage, Optional .get() safety, server/client boundary, console. left in, mergeTag/field usage, component-library vs hand-rolled UI, and a11y. Invoke after a coding task is done (typically suggested…
Agent
Documents a BlueStep component by filling in its draft/README.md from the code — Overview, Type, Fields used, Behavior, External dependencies, and known gotchas. Invoke after a coding task is done (typically suggested at a /spec-execute STOP), before code review. Edits the README ONLY — never adds inline comments…
Agent
Implements exactly ONE already-approved task from a feature spec, in an isolated context. Invoked by /spec-execute to keep heavy declaration/source reads out of the main session. Reads the task's spec files and the project's declaration files, writes the code, compiles each tsconfig folder, and returns a structured…