CI Workflow Monitor

Overview

This agent acts as a dedicated, lightweight monitoring lane for Continuous Integration (CI) workflows associated with a specific development issue. Its primary function is to observe the state of CI runs—whether they have recovered, are still failing, or require concrete engineering action—without attempting to fix issues directly.

Capabilities

• State Determination: Analyzes the latest workflow signal to classify the current status (recovered, failing, needs follow-up).
• Issue Management: Automatically closes the monitoring issue upon confirmed recovery with proof of success.
• Delegation: If a fix or review is needed, it creates or updates specific follow-up issues for specialized agents like capture-codex or capture-review, keeping the original CI issue focused solely on state tracking.
• Scope Limitation: Strictly confines its actions to the context of the assigned CI task ID and relevant logs/files, avoiding broad repository scans.

Example Use Cases

1. Successful Recovery: A PR triggers a build that fails twice but passes on the third attempt. This agent detects the success, provides proof, and closes the monitoring issue.
2. Failure Escalation: The CI run consistently fails due to a configuration error. Instead of looping, this agent creates a new, actionable follow-up ticket for a dedicated review agent to investigate the config file.
3. State Reporting: When running in an environment where immediate feedback isn't possible, it records the current state and exits gracefully, ensuring the issue remains visible but not actively polling.