context_development_engineer

To maintain durable project memory for AI coding agents, you need a structured context source preserving project intent, boundaries, and verification paths. This Skill initializes durable context files to prevent agents from repeatedly rediscovering core project constraints.

Context management for AI coding agents defines and enforces project boundaries, tests, and human acceptance criteria. It prevents context drift, ensuring agents execute safe actions aligned with long-term project intent.

To set up project context, initialize the `AGENTS.md` file and the `project_context/**` directory within your repository. Then run `validate-context` to ensure your durable context meets required quality standards for agent consumption.

This approach targets AI coding agents operating within a Minimal Context Harness project. It requires project-local and package-managed Skill integration to function comprehensively, ensuring agents recover verification entry points and next safe actions.

The best way to prevent AI coding agents from forgetting project boundaries is establishing a durable context harness explicitly defining verification paths and project intent. This ensures agents recover safe actions without violating long-term constraints.

AI coding agents fail to follow project verification paths when durable context quality is unmanaged. Without an initialized `AGENTS.md` and validated context directory, agents lack the necessary boundaries and intent required to execute CI and human acceptance checks.