What it does

Code Context Engineering is a high-precision pre-flight skill for AI agents that prevents "wrong-task" failures. Before a single line of code is written, it forces the agent to inventory disparate sources—including repository rules, user requests, existing codebases, external documentation, and historical chat memory—to identify contradictions, stale instructions, or missing criteria.

Why use this skill

Standard LLM prompting often results in "hallucinated intent" where agents guess how to resolve conflicting requirements. This skill replaces guesswork with a formal Conflict Gate. It enforces a "stop-and-ask" protocol if choosing a path would materially alter scope, safety, or data integrity. It builds a source-of-truth contract that ensures the downstream implementation is actually what the developer needs, not just what the agent inferred.

Supported workflows

• Complex Refactoring: Ensures existing behavior is preserved unless explicitly targeted for change.
• Feature Migrations: Detects version drift between installed packages and external documentation.
• Architecture Planning: Aligns current requests with long-term project rules and memory.
• Issue Triage: Identifies when ticket requirements are insufficient to begin coding.

The Output

The skill produces a Context Ledger with confidence labels and a Gate Status (Clear, Blocked, or Needs Verification). If blocked, it generates a high-signal, "low-AI-smell" clarification request that pinpoints the exact sources in conflict.