What problem does it solve?
Security posture drift in agentic infrastructure is hard to detect when threat modeling, MCP server scanning, and architecture alignment checks run as separate, uncoordinated commands with no historical record. This Skill bundles all three MetaHarness static surfaces into one timestamped audit record so drift over time becomes visible and schedulable.
Core Features & Use Cases
- Composite Audit Execution: Runs oia-manifest (L1-L9 layer alignment), threat-model (categorized MCP-surface threats), and mcp-scan (per-server policy and dependency findings) in a single pass.
- Worst-Severity Aggregation: Computes a composite worst severity from the threat model and MCP scan findings, with an --alert-on-worst threshold that exits non-zero when breached.
- Persistent Audit History: Stores each run in the metaharness-audit memory namespace under an ISO-timestamped key, enabling diff-based drift detection between audits.
- Graceful Degradation: Emits a standard degraded payload when MetaHarness components are unavailable, flagging partial results per component instead of failing.
- Use Case: Wire it into a weekly GitHub Actions cron job that fails the build whenever any HIGH-severity finding appears, while logging lower-severity drift for later review.
Quick Start
Ask the agent to run a composite MetaHarness audit on the current repository with an alert threshold of high severity and persist the result to the audit memory namespace.
