audit-context-building

The 'audit-context-building' skill is designed to enable ultra-granular, line-by-line code analysis for constructing deep architectural context before engaging in vulnerability or bug detection. It addresses the challenge of superficial code reviews where high-level summaries and assumptions can lead to missed edge cases or overlooked interactions between components. By enforcing a methodical, bottom-up approach, this skill allows Claude to build a precise mental model of the system, capturing invariants, assumptions, flows, and potential reasoning hazards. Its primary goal is to ensure that subsequent audit phases, such as security checks or bug identification, are grounded in comprehensive and accurate contextual understanding.

Key features include line-by-line and block-by-block code inspection, the application of micro-level First Principles reasoning, the 5 Whys, and the 5 Hows to every function or block of code. The skill continuously links insights from functions to modules and ultimately to the full system, updating its internal assumptions when new evidence is encountered. It also provides structured analysis formats, such as per-function microstructure checklists, and emphasizes the explicit documentation of invariants and dependencies to prevent context loss. The skill discourages speculative reasoning and ensures uncertainty is communicated clearly, promoting reliability and clarity in code comprehension.

This skill is particularly valuable for security auditors, software architects, and threat modelers who require deep understanding before vulnerability hunting or system analysis. It is intended for situations where bottom-up comprehension is critical, hallucinations must be minimized, and high-level guesses are insufficient. Typical use cases include auditing smart contracts, reviewing complex software architectures, and performing threat modeling where every line of code may impact system behavior. Users benefit from a methodical approach that reduces errors, enhances insight into system dynamics, and ensures that foundational understanding precedes any form of bug or vulnerability evaluation.