workflow-builder

Start a conversation describing your business process, and the builder guides you through seven phases: requirement profiling, phase definition, transition design, context schema, artifact design, Skill binding, and generation. It outputs a complete SKILL.md plus supporting schema and example files.

It generates SKILL.md as the main workflow instruction, status-schema.md defining the status.json structure, phase-examples.md with per-phase output examples, and optionally skills-reference.md when existing gdpa-cli Skills are bound to phases.

Yes. The transition design phase supports linear flows, conditional branches, skippable phases, rollback on failure, user confirmation gates, and loops with retry limits. These rules are encoded into the generated SKILL.md with ASCII flow diagrams.

Generated workflows use a status.json file stored under .gdpa/{session-id}/ to track phase status, context, and history. On restart, the workflow reads this file and resumes from the current phase, requiring the same --session-id across all gdpa-cli run commands.

Yes. Phase 6 of the builder lets you bind existing Skills such as devflow, edit-idl, bam-query, argos-query, iam, and tce to specific phases. The generated skills-reference.md documents each binding with usage examples.

Use workflow-builder when the fixed 7-phase base-workflow does not fit your process, such as database migrations, security audits, or multi-service integration testing. It lets you customize phase count, names, transitions, context fields, and Skill bindings.