How would you adopt these skills on a brownfield repo that's already full of docs from other workflows?

[bridgeyuwa]

Hi, and thanks for the skills.

My dilemma: I found these after the fact. The repo already has a huge pile of docs from other workflows: Superpowers-style specs and plans, lots of AI-generated discovery and research notes, and READMEs that contradict each other and the code. It's far too big to start over, and I'm not good at structuring docs (which is partly how it got like this), so tidying it by hand isn't realistic either. But I'd like to fix all of it and properly adopt these skills, not just use them on new work.

What I can't tell from the skill files is the intended path for a repo like this. Should I be converting and restructuring the old corpus into the shape the skills expect, or leaving it alone and having the skills use it as input to build the glossary, ADRs and tickets from? As far as I can tell, none of setup-matt-pocock-skills, grill-with-docs (via grilling and domain-modeling) or to-tickets say how docs outside the glossary/ADR locations are treated. So specifically:

1. Do old docs count as authoritative, as hints, or are they ignored unless I point at them?
2. If two docs disagree, or a doc disagrees with the code, is the idea that /grill-with-docs surfaces it and I decide before anything goes into GLOSSARY.md or an ADR?
3. Can I feed /to-tickets an old plan directly, or should I re-grill first? Superpowers plans are very code-level, so my guess is to re-slice from intent.
4. For someone who isn't good at docs, is the idea that the grilling does the structuring and I just answer and approve?

What I'm planning: move the old docs into a read-only legacy/ folder, add a line to AGENTS.md saying they're not authoritative and the code wins, then grill one area at a time and retire the old docs as I touch each area. The main thing I'm trying to avoid is a stale doc getting baked into the glossary or an ADR in the first session, since everything after would trust it.

Is that close to how you'd do it, or am I fighting the tools? If anyone has done this on a messy repo, I'd love to hear how it went.

A short video or README section on "adopting these skills on an existing messy repo" would help a lot of people, since most of the material I've seen shows the clean-slate path.