context-engineeringlisted

# Context Engineering Context engineering is the practice of curating the knowledge an LLM needs to work effectively in a codebase. Good context makes the difference between an LLM that follows project conventions and one that generates plausible but inconsistent code. Birgitta Boeckeler's key insight: the code design itself is a huge part of the context. Well-structured code communicates intent; the harness documentation makes that intent explicit and enforceable. ## The Context Section of HARNESS.md The Context section serves three audiences simultaneously: 1. **New team members** read it to understand the project's stack and conventions 2. **LLM agents** read it as structured input when reviewing code or generating enforcement 3. **The plugin** reads it to understand what tools might be appropriate for enforcement ### Stack Declaration Declare the factual baseline: languages, build system, test framework, container strategy. Keep this section factual — no aspirations, just what is true today. The `harness-discoverer` agent populates this during `/harness-init`. ### Conventions Conventions are the core of context engineering. Each convention must be **concrete enough that a reviewer could objectively say whether code follows it**. The test: could two independent reviewers agree without discussing it? For detailed examples of well-written vs poorly-written conventions, consult `references/convention-patterns.md`. ## Writing Effective Conventions ### The