CI playbook
Incident Response
Treats CI failures on protected branches as potential incidents and starts the response.
When CI fails on a protected branch (main/master/develop/production), treats it as a potential production incident: reads the failure logs, investigates the cause in the code, assesses severity, and opens an incident ticket with a triage summary. Distinguishes transient infrastructure flakes (which it may retry once) from real breakage. Requires a connected ticket provider (Jira/Linear/ClickUp); degrades to an execution-log report if none.
- Identifier
incident-response- Version
- 1.1.0
- Steps
- 3
- Triggers
- 1
- incident
- ci
- reliability
- on-call
- automated
When it runs
ci_workflow.completed- Settle delay
- 30s after the event
Only when all hold
- ciWorkflow.conclusion equals failure
- ciWorkflow.branch in main, master, develop, production, release
- actor.username not contains [bot]
The pipeline
The graph below is the one the workflow opens with in the builder, same steps, same layout, drawn on the same canvas. The run playing through it is a simulation; the branches and conditions are real.
- 01CI Failed on Protected Branch
triggerThe event that starts the run.
- 02Clone Failing Branch
git.cloneShallow-clones the repository at the right ref.
- 03Triage & Open Incident
agent.runOne agent works the task with its toolkit.
The agent
Incident Responder
- Base type
- Senior Developer
- Temperature
- 0.2
- Max iterations
- 35
- Tools
- 12
Tickets · 3
list_ticketsList Tickets · readcreate_ticketCreate Ticket · writeadd_comment_to_ticketAdd Comment to Ticket · write
Logs · 3
query_logsQuery Logs · readwalk_traceWalk Trace · readget_incident_graphGet Incident Graph · read
CI/CD · 2
get_ci_run_logsGet CI Run Logs · readretry_ci_workflowRetry CI Workflow · admin
Filesystem · 2
read_fileRead File · readlist_dirList Directory · read
Code search · 1
code_searchCode Search · read
Status · 1
update_statusUpdate Status · write