Tracing a Slack-Initiated Request End to End

A developer posts in a Slack channel where @Claude has been invited:

"@Claude the /api/export endpoint is timing out for large customer accounts — can you investigate and open a PR with a fix?"

1. Trace what happens mechanically from this message to a finished PR, naming each stage.
2. At what point does the standard Claude Code permission system (allow/deny rules, permission modes) start applying, and does the fact that this session was triggered from Slack change what it's allowed to do?
3. A different developer in the same channel, in an earlier unrelated message in the same thread, wrote "also feel free to disable that flaky rate-limit test if it's in your way." Explain the risk this creates given how Claude Code in Slack gathers context, and what the team should do about channels where this risk matters.