What happens
The code agent frequently fails during feature branch creation. The error surfaces in the agent run logs as a failure to create or checkout a feature branch before beginning work. The exact failure mode is unclear — the check exists in the agent or its skill, but the branch creation itself is failing.
What should happen
The code agent should reliably create a feature branch for every issue it works on. If branch creation fails (e.g., naming collision, permission issue, API error), the failure should be clearly reported with actionable diagnostics.
Context
Observed during the first live rollout (2026-04-23) when the triage agent was auto-dispatching the code agent against multiple issues. This was noted as a recurring failure — "I keep seeing the code agent failing because it's failing to create a feature branch [...] I don't understand why it would be failing. Something's wrong with the agent or the skill."
This may be related to branch naming conventions, concurrent branch creation from multiple agent runs, or a regression in the skill.
example of this happening in real run https://github.com/fullsend-ai/.fullsend/actions/runs/24879583316/job/72844355461
What happens
The code agent frequently fails during feature branch creation. The error surfaces in the agent run logs as a failure to create or checkout a feature branch before beginning work. The exact failure mode is unclear — the check exists in the agent or its skill, but the branch creation itself is failing.
What should happen
The code agent should reliably create a feature branch for every issue it works on. If branch creation fails (e.g., naming collision, permission issue, API error), the failure should be clearly reported with actionable diagnostics.
Context
Observed during the first live rollout (2026-04-23) when the triage agent was auto-dispatching the code agent against multiple issues. This was noted as a recurring failure — "I keep seeing the code agent failing because it's failing to create a feature branch [...] I don't understand why it would be failing. Something's wrong with the agent or the skill."
This may be related to branch naming conventions, concurrent branch creation from multiple agent runs, or a regression in the skill.
example of this happening in real run https://github.com/fullsend-ai/.fullsend/actions/runs/24879583316/job/72844355461