This assumes the population is equally distributed in both vaccinated and unvaccinated groups, such that the amount of people that could die of covid in each group is equal to the infection fatality risk (its value isn't relevant). As far as I know this a premise of all randomized vaccine studies.

Some definitions:
Vd% = percentage of total population already vaccinated in a given day
NVd% = percentage of total population still non-vaccinated in a given day

Vdd = Absolute number of vaccinated that died of COVID-19 in a given day
NVdd = Absolute number of non-vaccinated that died of COVID-19 in a given day

The idea:
If vaccines are useless, the distribution of already vaccinated in a given day should be the same as the distribution of vaccinated that died of COVID-19 in that given day, that is

Vdd / (Vdd + NVdd) = Vd%

Consequently

NVdd / (Vdd + NVdd) = NVd%

If vaccines are worse than worthless, that is, are causing harm

Vdd / (Vdd + NVdd) > Vd%

and as such

NVdd / (Vdd + NVdd) < NVd%

An example:

On 22 July, 2021, for Brazil
>Vd% = 46%
>NVd% = 54%
https://ourworldindata.org/covid-vaccinations

>Vdd = 168 + 123 + 37 = 328
>NVdd = 265
https://zerobias.info/

As such
Vdd / (Vdd + NVdd) = 328 / 593 = 55.31%
NVdd / (Vdd + NVdd) = 265 / 593 = 44.69%

Compared to the Vd% of 46%, 55.31% died
Compared to the NVd% of 54%, 44.69% died

So, 22 July 2021, if you were a brazilian, you had a 18.62% greater chance of dying of SARS if you were vaccinated compared to the unvaccinated.

If vaccines are causing harm, I'd not to assume an interval of 2 weeks to consider anyone fully vaccinated, but if you want to do that for a more accurate analysis, you could use the number 2 weeks before the given day for Vd% and NVd%, and the given day numbers for Vdd and NVdd.