Names the domain (logical/semantic bugs, program intent) and describes the general action (detects bugs, catches issues), but doesn't list multiple specific concrete actions like 'trace variable states', 'verify loop invariants', or 'check boundary conditions'.

Clearly answers both what ('Detects logical and semantic bugs by understanding program intent') AND when ('Use when static analysis has already run and found nothing, when the user reports incorrect behavior but no crash, or when reviewing algorithmic code for correctness').

Includes natural keywords users would say: 'logical bugs', 'semantic bugs', 'incorrect behavior', 'no crash', 'algorithmic code', 'correctness', 'static analysis'. These cover common variations of how users describe this problem.

Carves out a clear niche distinct from syntax checkers, linters, and static analysis tools by explicitly positioning itself for cases where those tools fail. The triggers ('static analysis found nothing', 'incorrect behavior but no crash') are highly specific.

Every section earns its place—no explanation of what bugs are or how Python works. The signal catalog table is dense with actionable patterns, and the worked example demonstrates rather than explains.

Provides a concrete signal catalog with specific patterns to look for, a complete worked example with real code and test confirmation, and a clear output format. The intent hierarchy and FP suppression rules are immediately applicable.

Clear sequence: establish intent → detect smells → rank by confirmability → suppress FPs → output findings. The worked example demonstrates the full workflow including the critical validation step of writing a confirming test.

Well-organized with clear sections (intent question, signal catalog, ranking, FP suppression, worked example, edge cases, do not, output format). Appropriate length for a single file with no need for external references.