Hi everyone,
I’m using Cursor Cloud Agents (triggered from Slack automations) on a team repo that has strict conventions for branch names and git history.
I have two related questions:
1. Branch prefix: can it follow feat/, fix/, etc. automatically?
Our project convention is something like:
feat/<short-description>fix/<short-description>hotfix/<scope>
But Cloud Agents always create branches like:
cursor/mkraft-orders-screen-improvements-d0dc
I added explicit rules in our AGENTS.md / .cursor/rules/agents.mdc telling the agent to follow our branch naming convention, but the agent still starts on a cursor/... branch generated by the platform.
From what I’ve read in other threads, the Cloud Agent dashboard lets you change the default prefix (e.g. from cursor/ to something else), but:
- If you leave the prefix empty, it falls back to
cursor/ - There doesn’t seem to be a way to disable the prefix entirely
- The platform prefix appears to override repo-level rules in
AGENTS.md
Question: Is there a supported way to make Cloud Agents automatically use conventional prefixes like feat/ or fix/ based on the change type (feature vs bugfix), ideally respecting rules in AGENTS.md? Or is the only option a single static prefix in the dashboard (e.g. always feat/)?
2. Commit author: can commits be attributed to me instead of Cursor Agent?
When the Cloud Agent commits and pushes, the author shows as:
Cursor Agent <[email protected]>
For our workflow (auditing, PR ownership, team conventions), we’d prefer commits to appear under the GitHub user who triggered the agent session, not the generic Cursor Agent identity.
I understand there may be a Co-authored-by: trailer for the session initiator on hosted agents, but we’d like the primary author to be the person who started the run.
Questions:
- Is it possible to configure Cloud Agents (hosted) so commits use the initiator’s
user.name/user.email(or GitHub noreply email) as the author? - If not, is self-hosted Cloud Agent the only way to control
git config user.nameanduser.emailon the worker? - Is there any recommended workflow (squash merge, cherry-pick, amend, etc.) that teams use to keep attribution correct without manual rework every time?
Environment
- Cursor Cloud Agent (hosted), triggered via Slack automation
- GitHub repo with
homolog/masterworkflow - Rules in
AGENTS.md/.cursor/rules/agents.mdc(branch naming, Conventional Commits, etc.)
Any guidance or planned improvements on branch naming flexibility and commit attribution would be very helpful. Thanks!