os-what-could-go-wrong

1# os-what-could-go-wrong
2 
3Assume it already failed, then work backwards to find out why - while there is
4still time to change it. The attack runs in a fresh agent that had no part in
5the decision, because an agent that helped shape one reviews it far too
6gently: it defends its own reasoning, and it misses the thing that kills the
7plan out of politeness.
8 
9`os-ask-simple` screens a choice before the user picks one. This one runs
10after the choice is made and before it can no longer be taken back.
11 
12## Language
13 
14Write in the language the user speaks in this session, detected from the
15conversation. Names, figures and identifiers stay as they are. The fresh
16agent cannot see this conversation, so the language travels in the handover.
17 
18## Step 1 - write down what is actually being decided
19 
20Find the decision first. Given as text or a document, that is it. Asked at the
21end of a discussion, it is the decision the discussion arrived at, and say
22which one you took it to be. If neither, ask which decision to attack.
23 
24Then fill every line. This is the only thing the fresh agent will ever see.
25 
26```
27DECISION BRIEF
28What will be done: three to seven sentences
29What it is for: the problem it solves, and what success looks like, measured
30  wherever it can be
31The main moves: the money, the people, the systems, the dates
32What is fixed: the constraints, plus the surrounding facts that matter
33Who it lands on: who and what is affected if this goes wrong
34What cannot be undone: which parts are one-way
35…