I'm not sure whether or not we ever had consensus on how to manage access to Corto. For example: which channel(s) the bot will ultimately reside in, who is able to issue commands, which commands, and how we go about ensuring that a nick on IRC corresponds to the IRL person we trust. However, the design doc says:

Public and private incident views

We would like to operate in the public as much as possible, however many incidents must be private to avoid aiding our adversaries, and initially keeping > information private by default is generally the safest path. How can the tooling help with these conflicting objectives?
For the first iteration all Phabricator tasks will be private by default. This will ensure that responders are not concerned about sensitive information in the > Google document being transferred into the Phabricator task.

In addition to creating issues private by default, presumably the bot will (initially) only reside in #mediawiki-security? If so, is it likewise safe to assume that every member of #mediawiki_security is able to run every (incident-related) bot command? If not, in what way is access structured?

Summary:

For purposes of the MVP:

• The bot will reside only in #mediawiki_security, and will apply no access control (any user of the channel can run every command)
• Phabricator tickets will be created with restricted view & edit policies (T376500). Incidents that can be made public will be made so via the Phabricator web interface after-the-fact.