h4tcher-scoped-audit

To run a read-only quality audit on a codebase slice, you provide a natural-language scope description like a single file or subsystem, which the tool resolves into a concrete file list to analyze without modifying your repository.

Yes, you can scope a codebase analysis to a small group of files by describing the target area in natural language, which triggers a bounded review that resolves file paths via globbing and applies a fresh 4–8 item checklist.

A scoped audit generates severity-graded findings by applying a fresh rigor-contract checklist to the resolved files, synthesizing causal analysis with confidence levels and pillar traceability into a concise in-chat report.

No, the scoped audit workflow does not write changes to your repository because it operates under strict read-only constraints, only emitting sub-agent artifacts internally and writing to a designated `.audit-workspace/` directory.

The best way to audit a subsystem without a full 21-domain review cycle is using a bounded scope audit, which creates a targeted 4–8 item checklist and optionally executes parallel sub-agents to produce a focused 80-line report.

To include critical files when defining scope for a repository slice audit, you mention them directly in your natural-language prompt, allowing the globbing logic to resolve them alongside the described subsystem boundaries.