i ran a model checker on one grok bot yesterday. it found real holes
Run a model checker over your bots' rules: on 12 bots, 34 of 39 rules failed, including bots that would take another bot's "nick said yes" as approval.
For
Your bots act on each other's messages and you haven't tested the rules.
When you're done
34 of 39 rules failed, each check under a second, now being fixed.
How
- 01
Use /statechart from @poteto's pstack to model-check a bot's rules.
- 02
Run it across your other bots.
- 03
Review what failed (here, 34 of 39 rules).
- 04
Fix bots one at a time, approving every fix.
Notes
What's good: it caught an inbox bot that would trash email because another bot said OK. The catch: the author approves every fix himself, one bot at a time.
Related plays
Write-ups follow public posts and official docs. No likes, views, or dashboard numbers. Open the source to check the original. About