Harness Guards

Harness guards are context-aware safety constraints that gate phased plan execution and protect knowledge artifacts. They work by scanning project DNA and risk profiles to inject mandatory governance rules into specific workflow steps and file headers.

The best way to enforce policy compliance is using a centralized guard catalog that defines exact syntax and escalation patterns. This ensures an AI agent follows the right governance rules at the correct moment without relying on generic warnings.

You generate harness guards by loading a guard catalog and scan protocol, which scans your project rules and risk mitigations. The system then emits correctly formatted HTML comment guards tied to specific phases, tasks, and file headers.

Yes, you can use automated safety gates to protect critical knowledge artifacts during creation and editing. The system applies progression phase rules that gate side-effects and enforce checkpoint actions before modifying files.

No external dependencies are required, but the system needs to load guard-catalog and scan-protocol references. These sidecar references provide the data necessary for just-in-time guard generation while keeping logic out of the skill itself.

Generic warnings fail because they lack context-specific gating tied to your project's unique risks. Context-aware harness guards solve this by scanning rule indices and risk tables to produce plan-specific, constraint-specific safety barriers.