Find the patterns worth reviewing.

Find repeated database or network calls, file reads, subprocesses, parsing, serialization, regex compilation, sorting, and selected expensive nested operations inside loops.

A match identifies where work repeats. Runtime impact still depends on call frequency, input size, caching, and the surrounding execution path.

Review dynamic evaluation, Python pickle deserialization, suspicious exception handling, and selected concurrency or initialization patterns that can hide failures or create surprising behavior.

Deep analysis adds supported cross-file and data-flow checks where the analyzer can establish the required identity and path evidence.

Structural metrics surface cognitive and cyclomatic complexity, long methods, and oversized objects. Test checks look for missing or low-value assertions. Credential checks identify suspicious locations without reproducing matched values.

Reports preserve check status and coverage. Unsupported languages, unavailable models, skipped files, time limits, and incomplete discovery remain visible rather than becoming a misleading clean bill of health.

• 20 quick-review check families
• 27 deep-review check families
• Traceable locations and bounded excerpts

• Repository analysis Understand review scope, inputs, and safety boundaries.
• Custom static analysis Add checks based on the rules your team already follows.
• Sample report See how a finding connects to exact source evidence.