Guidelines for using agents with Sigrid

Each of these guides follows one job through on a real codebase: what to configure, what a session looks like, and how to check what the agent did. They use Sigrid Axis, so install it first.

• Building with Guardrails puts Guardrails in the feature loop, so the agent checks each file it writes before you see the diff.
• Preventing architecture drift checks an agent’s diff against Sigrid’s measured dependency graph before it merges.
• Improving architecture finds the directory whose structure is most worth fixing, plans the move or facade that fixes it, and checks the result against Sigrid’s graph.
• Reducing technical debt works through the refactoring candidates that carry the most weight in your maintainability rating.
• Triaging security and reliability findings classifies a findings backlog against how your system is deployed, records a rationale for every decision, and fixes what should be fixed.

These are workflows we run ourselves. They do not cover everything Axis can do, so for the rest, see the skills reference, the MCP tools reference, and configuration.

Choose a model

Pick the model by how much judgment the work needs. That depends on your codebase as much as on the skill: triage is routine when the deployment is simple, and untangling a directory that everything depends on can take real reasoning.

Each vendor offers models in three tiers:

Each guide suggests a starting point in a block like this one. Adjust it to your codebase.