GitHub Project Deploys AI Bug Triage & Fix Workflows with Safety Guardrails

A new GitHub project introduces a structured, tool-agnostic framework for automating software bug management using AI. The core innovation is a dedicated `.agents/` directory containing workflows designed to triage Jira issues and execute code fixes autonomously. The system is built with explicit safety mechanisms to prevent uncontrolled AI actions, including a circuit breaker that halts operations after three failed fix attempts, a strict one-issue-at-a-time policy, and mandatory human-in-the-loop (HITL) checkpoints. It also features automated cleanup procedures to revert changes if a workflow fails.

The framework is designed for progressive context disclosure, using a hierarchical `AGENTS.md` file structure so any AI tool can access relevant operational guidelines based on its working directory. It comes packaged with extensive reference documentation covering a label taxonomy, Jira Query Language (JQL) templates, bug categorization, an ecosystem dependency map, standardized comment templates, and known fix patterns. A key prerequisite is a `/setup-preflight` utility that verifies all necessary tool access and permissions before any automated process begins, aiming to prevent runtime failures.

Notably, the system includes container-aware assessment guidance, instructing agents to inspect lock files and container images rather than local virtual environments, and provides registry reference docs for querying the Red Hat container catalog via the Pyxis API. A utility script for extracting content from Google Docs is also included, suggesting integration with external documentation sources. The project status indicates core components like reference docs and the AGENTS.md hierarchy are complete, positioning it as a ready-to-implement system for teams seeking to integrate AI into their software maintenance lifecycle with controlled risk.