Methasit-Pun Data Project Lifecycle (Router)
This is a router skill covering a data project from "we have an objective" to "the code is clean." Figure out which stage the user is in, then invoke the matching stage skill with the Skill tool. A greenfield project runs the stages in order; a mid-flight project jumps to its stage.
The lifecycle
| Stage | Question it answers | Invoke sub-skill |
|---|---|---|
| 1. Discover | What data do I need and where do I get it? | data-sourcing |
| 2. Profile | What does the data actually look like? (checks + ER) | data-profiling |
| 3. Architect | How should the whole platform be structured, on what infra, at what cost? | data-architecture |
| 4. Build | Design the bronze/silver/gold pipeline, gold-first | medallion-design |
| 5. Refactor | Make the resulting notebook/code reviewable | notebook-refactor |
Routing rules
- "New project / where do I start" → begin at stage 1 (
data-sourcing) and proceed in order, confirming each stage's output before advancing. - User already has data in hand → start at stage 2 (
data-profiling). - Sources + shape known, needs a plan → stage 3 (
data-architecture). - Architecture decided, ready to build layers → stage 4 (
medallion-design). - Working code that's messy → stage 5 (
notebook-refactor) — can run anytime, independent of the others. - Invoke by name, e.g.
Skill(skill="data-architecture"). For a task spanning stages, invoke each in sequence and carry the output forward.
How this router relates to the others
The lifecycle produces work that the topic routers deepen:
- Stage 3
data-architecturedelegates cloud + cost to → [[cloud-data-infra]] - Stage 4
medallion-designdefers pipeline mechanics to → [[data-pipelines]] and the model to → [[data-modeling]] - Profiling & building both hand validation/governance to → [[data-reliability]]
- ML objectives branch to → [[ml-feature-engineering]]
Use this router for "what stage am I in"; use the topic routers for "go deep on pipelines/modeling/reliability/cloud."