hello-auditkit

Auditing AI assistant configurations involves analyzing prompts, SKILL.md files, plugins, MCP/LSP components, and memory files to produce verifiable findings and fix recommendations. This Skill evaluates configuration quality and safety-by-default behavior using strict prompting constraints and evidence-grounded line anchoring.

A compliance check on memory files like AGENTS.md or CLAUDE.md requires type-aware auditing that detects specific configuration types and generates verifiable findings. The audit evaluates safety-by-default behavior and prompting quality, outputting evidence-grounded fix recommendations saved directly to a file.

Yes, you can audit pasted text directly, as the configuration audit supports single files, directories, or pasted text inputs. The Skill detects the configuration type from the provided content and applies safety checks and prompting quality validation to generate a file-based report.

No, the configuration audit follows a safe-by-default workflow that keeps scope limited to user-provided materials and requires explicit confirmation before applying any fixes. It uses a staged flow with confirmation gates to ensure you review evidence-backed findings before modifications occur.

The AI assistant audit outputs reports to a file to keep terminal output minimal and improve long-review workflows. This file-only reporting approach ensures verifiable findings, line-anchored evidence, and fix recommendations are saved for compliance-style checks and future reference.