addy-context-engineering

Prevent AI output quality decline by enforcing hierarchical context loading of rules, specs, and relevant files. This deliberately curates persistent and task-scoped context to maintain predictable agent behavior and guardrails when switching between tasks or projects.

Start a new development session with project conventions applied automatically by creating a project-wide AGENTS.md rules container. Loading this as persistent context enforces conventions and boundaries across all sessions for consistent behavior.

Agent output quality declines when switching between tasks or projects due to lost hierarchical context layering. Applying persistent rules plus per-feature specs and relevant file references adapts the agent to changing work scopes and restores predictable behavior.

Hierarchical context loading organizes rules, architecture docs, and relevant source files from persistent project-wide context down to task-scoped context. This rules-driven guidance supports per-feature specs and file references to adapt the agent to changing work scopes.

You need an AGENTS.md or equivalent rules container to enforce coding conventions and boundaries across AI sessions. Loading it as persistent context provides rules-driven guidance that establishes predictable behavior and guardrails for the agent.

Layer context for specific features by supporting persistent project-wide rules alongside per-feature specs and file references. This context layering adapts to changing work scopes while maintaining conventions and predictable agent behavior across sessions.