Design principles

Few assumptions about the Hunt -- we won't know much it about in advance.
1. There are puzzles.
2. Every puzzle has a word-or-short-phrase solution.
3. Every puzzle has a unique name.
4. Puzzles are probably divided into rounds, but that may be a little wonky (e.g. in 2018 some puzzles belonged to more than one round).
5. Every round probably has a metapuzzle.
Easy to stop using.
1. If the tool does the wrong thing, we should be able to correct it easily.
2. If the tool doesn't work, e.g. because an assumption is wrong, we should be able to immediately go back to doing things by hand without losing anything.
3. If any individual QM doesn't want to use the tool, they should be able to do QM stuff by hand without the tool getting in their way or them getting in the tool's (as long as it's not the same puzzle at the same time).
No actions that can't be undone.
1. Sometimes deleting a row or a sheet is the right thing, but those cases are unusual and we'll do them by hand.
The tool doesn't need to be secret or pretend to be human, but it should be polite.
Nobody should have to care about it except the QMs.
1. Solvers shouldn't need to learn how to interact with the tool -- it should conform to them, not the other way around.

Design decisions

No or minimal data schema. We won't try to model the Hunt. [1]
No data store except the human-readable and -editable Google Sheet. [1, 2]
Every action is manually triggered -- we won't try to monitor the Hunt site for changes. (If we feel really clever we might change this to "manually confirmed" but that's as far as we go.) [2]
No automated deletion or destruction of anything, either when a puzzle is solved or when it was added by accident. Reversible archiving is okay. [3]
Automated messages are allowed but should never be annoying or spammy. [4, 5]
We won't integrate at all with the content of the spreadsheets used to solve puzzles. [5]
No tool-assisted state updates before a puzzle is solved (like from "Working" to "Stuck-E") -- solvers or QMs will update the tracking sheet directly. [5]
No change to the "I have an answer, call it in" workflow; QMs just use the tool to update state after they get the phone call confirming the answer was correct. [5]