Shared read-only discovery, review, and reporting protocol for every Sarj judgment-layer audit command.
Scanned 9/23/2026
Install to Claude Code
npx -y skills add sarj-ai/code-standards --skill audit-protocol --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Audit Protocol?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/sarj-ai-audit-protocol)More formats (shields.io, HTML) on the badges page.
---
name: audit-protocol
description: Shared read-only discovery, review, and reporting protocol for every Sarj judgment-layer audit command.
---
# Audit protocol
Every audit command follows this read-only workflow unless the user separately requests implementation. Domain commands contain only judgment checks. Deterministic checks are centralized in `code-standards`; do not duplicate their findings in an audit report.
## Discover
- Inspect manifests, lockfiles, source roots, generated boundaries, framework configuration, database dialect, and native lint, type, and test commands.
- Run deterministic checks first and verify their findings in source.
- Gate checks by the detected stack, including database, runtime, rendering, and tenancy differences.
## Review
Inspect only the selected command's concerns across applicable source roots. Exclude vendored, generated, fixture, snapshot, build, and lock files unless the command explicitly targets them.
## Report
Order findings by severity and confidence. Each finding includes its file and line, evidence and impact, the smallest credible remediation, confidence, and its supporting deterministic rule or tool, or `judgment-only`.
Separate confirmed defects from suggestions. Omit speculative, stylistic, duplicate, and unverified tool findings. When no findings remain, state the scopes and checks examined.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!