Agent Kit by AtlasScope
GitHub

Prompt·Data integrity and analytics·D06

Code health, with proof

A general health check of the code that refuses findings without evidence.

Markdown
Prompt
Perform a focused code-health audit calibrated to this repository's own conventions and operational risks. Require a specific file/line, reproducible command, trace or commit for every finding; grade confidence and impact; reject generic style complaints. Return a short ranked list of repairable issues rather than hundreds of speculative observations.
What it returns
A ranked list of under fifteen issues, each with file:line or a command, confidence and impact.
Skip it when
When a narrower investigation above already fits the symptom.
Changes
Changes nothing
Sessions
One session
Tools
Repository access.
Also called
Evidence-required code-health audit

Skills it uses

harsh-reviewUse only when Austin explicitly invokes /harsh-review or $harsh-review. Gives a blunt, severity-ranked critique of code or product work, backed by evidence at every item, then ships the small certain fixes in the same turn and verifies them. Ordinary requests for honesty, feedback, or a review must not activate it.working-style·Agent Kit

Related prompts

More in Data integrity and analytics.