Full-Codebase Audit FAQs

Question 1

How does this differ from standard AI code reviews?

Accepted Answer

Most AI tools use RAG or chunking which loses context between files. This skill loads the entire project into memory, allowing Claude to find bugs that span multiple files, such as auth bypasses or architectural violations.

Question 2

Can I run a security-only audit?

Accepted Answer

Yes, the skill supports arguments like '/ork:audit-full security' or '/ork:audit-full architecture' to focus the analysis and save on token usage.

Question 3

What is the 1M context window requirement for this skill?

Accepted Answer

To perform a full-codebase audit, you generally need a Tier 4+ Claude account and Opus 4.6 to access the 1M token window. This allows the skill to load roughly 100K lines of code for cross-file analysis.

Question 4

Which programming languages are supported?

Accepted Answer

While it can read most text-based code, it is highly optimized for TypeScript, JavaScript, and Python, featuring specific patterns for FastAPI, React 19, and LangGraph.

Question 5

What happens if my codebase is too large for the context window?

Accepted Answer

The skill includes an estimation script to check your token budget. If the project is too large, it will suggest narrowing the scope to specific directories or falling back to the /ork:verify skill for chunked processing.

Full-Codebase Audit

Key Features

Use Cases

Full-Codebase Audit

Key Features

Use Cases