I don't know about killing the agents, but there have been known examples of that kind of false flag terrorist attack. Probably the most famous is the Lavon Affair, but there's also:
There have also been a number of false-flag incidents in which a government attacked its own people with terrorism, but, whether by luck or by intention, no-one died:
There are also various well-known incidents that look like false flag terrorist attacks by a government on its own people, but which are disputed:
It is also possible that there were other such incidents but where the false-flag activity went undetected.
Overall, such false-flag attacks seem to be extremely rare, but not unknown. They seem to normally occur in the context of an existing violent conflict. As we have no particular evidence that the Charlie Hebdo attacks were a false-flag operation, our prior that the French government was involved should be extremely low.
After the terrorist attacks at Charlie Hebdo, conspiracy theories quickly arose about who was behind the attacks.
People who are critical to the west easily swallow such theories while pro-vest people just as easily find them ridiculous.
I guess we can agree that the most rational response would be to enter a state of aporia until sufficient evidence is at hand.
Yet very few people do so. People are guided by their previous understanding of the world, when judging new information. It sounds like a fine Bayesian approach for getting through life, but for real scientific knowledge, we can't rely on *prior* reasonings (even though these might involve Bayesian reasoning). Real science works by investigating evidence.
So, how do we characterise the human tendency to jump to conclusions that have simply been supplied by their sense of normativity. Is their a previously described bias that covers this case?