Developer Coding Surfaces

Developer Coding Surfaces are three composite tools designed specifically for coding agent workflows. They address the three questions coding agents ask most often that existing tools don't answer directly: "what went wrong last time we touched this?", "why is it built this way?", and "is it safe to deploy?"

The incident context tool (get_incident_context) retrieves past incidents, deploy failures, gotchas, and active constraints for a module or system. When a coding agent is about to modify infrastructure code, deploy scripts, or any module with a history of fragility, this tool surfaces the tribal knowledge that prevents repeat failures. Docker bind mount traps, Vault agent atomic-write patterns, stale dist/ artifacts, port conflicts - the kind of operational landmines that only exist in people's heads or scattered across conversation history. The tool composes memory queries (both broad and recency-weighted), constraint lookups scoped to the module, and domain changelog entries into a single ranked response.

The architecture rationale tool (get_architecture_rationale) aggregates all architectural decisions, active constraints, corrections, and skills for a domain into a coherent narrative. Coding agents waste enormous context asking "should I refactor this?" when the answer is often "it's built this way because of compliance requirement X" or "we tried the alternative in March and rolled it back because of Y". This tool prevents unnecessary refactors of intentional design choices by surfacing the full decision history, including corrections that show what was tried and abandoned.

The deploy readiness tool (check_deploy_readiness) cross-references changed modules against active constraints, recent incidents, knowledge freshness, and active alerts. It returns a composite verdict - ready, caution, or block - with a per-module breakdown showing exactly which constraints matched, how many recent incidents exist in the same area, whether knowledge is stale and needs verification, and what active alerts might interact with the changes. This is the pre-deploy safety net that catches constraint violations and known fragile paths before they reach production.

All three tools are read-only composite projections - they compose existing memory, constraint, and alert data into developer-focused views without adding new storage or mutations.