What problem does it solve? When building LLM-powered coding agents, uncontrolled growth of the model context causes cache misses, runaway token costs, and degraded inference quality. This Skill gives reviewers a concrete checklist for auditing how context items are injected into model requests. ## Core Features & Use Cases - Context Integrity Rules: Enforces incremental context building with no history rewrites and no frequent changes that cause cache misses. - Bounded Size Enforcement: Requires every injected item to have a hard cap, with no single item exceeding 10K tokens and items over 1K tokens flagged as P0 for manual review. - Structured Fragment Requirement: Mandates that all injected fragments be defined as structs in core/context implementing the ContextualUserFragment trait. - Use Case: While reviewing a pull request that adds a new system reminder or tool output to the agent's prompt, apply this Skill to verify the addition is bounded, cache-friendly, and implemented as a proper context fragment. ## Quick Start Review this pull request's changes to model context injection using the code-review-context checklist.
