The giveaway
The starter skoorboard.
Fill this in by hand, on any stack, without us. In about two minutes you will know exactly what you can — and cannot — answer about your own agents. The blanks are the point.
You do not need a Skoor account to keep a skoorboard. Keep this one in a spreadsheet if you like. What Skoor adds later is measurement — a score computed against a published rubric, because a firm cannot independently attest to itself.
The roster
List every AI agent running at your firm, and who owns each one.
If you cannot name them all, that is the finding. An agent nobody can name is an agent nobody governs.
The scopes
For each agent, what systems can it reach — and what can it never do?
A permissions list with no refusals is half an answer. The refusals are the part an examiner asks about.
The trail
Where does the record of what each agent did actually land — and is it complete?
If any row is blank, you cannot answer 'what did it do?' for that agent. Blank is honest; it is also the gap.
The scores
For each agent, what is its Company Skoor / Agent Skoor today?
Empty slots are the correct state. This is where a measured Skoor goes once one exists.
What your blanks are telling you
- Blank roster — you do not know what is running. Start there.
- Scopes without refusals — you can say what an agent does, not what it can never do. That gap is what a client or examiner probes first.
- A blank trail row — for that agent, you cannot answer “what did it do?” There is no record to point to.
- Empty score slots — correct. Nobody has measured them. A number here that you did not compute would be the exact thing you are trying to avoid.
Every blank you just found is a question someone will eventually ask you. That is what an assessment turns into a measured baseline — and what a Skoorkeeper keeps current after.
Turn the blanks into a baseline.
The assessment produces your first real Company Skoor — two weeks, fixed fee, and you keep the report either way.