[n1xan]

This pull request introduces and refines the agent prompts for the BELLATRIX agentic QA framework, clearly defining the roles, procedures, and boundaries for each specialist agent. The changes ensure that each agent operates with explicit instructions, enforces critical thinking and stakeholder engagement, and formalizes a continuous improvement loop via the Agent Doctor. The most important changes are summarized below:

Agent Roles and Procedures

• Added a new .claude/agents/agent-doctor.md file, defining the Agent Doctor’s responsibilities for reviewing agent performance after each delivery, identifying instruction weaknesses, and proposing improvements, with a structured review and application process.
• Refined the .claude/agents/qa-coordinator.md to clarify the coordinator’s gate-based workflow for QA requests, emphasizing critical thinking, stakeholder clarification, explicit plan approval, and the proper sequencing and handover between agents. Added detailed instructions for capturing stakeholder feedback and integrating the Agent Doctor review into every delivery.

Specialist Agent Guidance

• Updated .claude/agents/automation-engineer.md to provide step-by-step procedures for implementing page objects and tests, including exploration before coding, strict adherence to test cases, and explicit instructions for handling failures and what not to change.
• Updated .claude/agents/test-analyst.md with detailed steps for requirements analysis, application exploration, test case design, ambiguity handling, and reporting of potential issues, ensuring that test design precedes any automation and aligns with stakeholder expectations.

Governance and Safety

• Across all agent prompts, reinforced strict boundaries on what agents may and may not change (e.g., never editing application code or committing changes without approval), and formalized the process for updating instructions only with stakeholder consent. [1] [2] [3] [4]