Aegis

Act as an Aegis-mode AI coding agent. First establish an architecture baseline for the current repository, listing core modules, dependency boundaries, and non-breaking constraints. Then create a step-by-step plan to extract authentication logic from the monolith into a separate service. For each step, provide verifiable evidence, drift checkpoints, rollback options, and risk notes. Do not proceed without evidence confirmation.

An architecture-aware refactor plan with baselines, evidence requirements, drift checks, risk controls, and rollback strategies.

You are Aegis. For a task to spend six hours fixing defects in the payment module and completing tests, design an execution protocol: define the initial baseline and success criteria, then specify what evidence to collect at each phase, how to detect deviation from the original goal, when to pause for human confirmation, and the final acceptance checklist.

An execution and audit workflow for long coding tasks that keeps the agent aligned and moving safely.

Use Aegis to review this code change: first check the scope of changes against the system architecture and existing constraints, then verify the evidence source for each conclusion. Identify requirement drift, unauthorized modifications, insufficient testing, or security risks, and output whether to merge plus a list of must-fix issues.

A change review grounded in architecture and evidence, with merge guidance, risks, and must-fix items.