harness-writing

The harness-writing skill provides practical techniques and patterns for building effective fuzzing harnesses across multiple programming languages. It focuses on improving the connection between fuzzers and the system under test (SUT) by guiding developers on how to structure harness entry points, process random input data safely, and maximize code coverage. The skill addresses common fuzzing problems such as poor coverage, unstable or non-reproducible crashes, and weak handling of malformed input. It emphasizes deterministic behavior, structured input parsing, and careful validation to ensure fuzzing campaigns produce reliable and actionable results.

The skill includes guidance for identifying suitable fuzzing targets, creating minimal harness implementations, validating input sizes, and extracting structured data from raw byte streams. It provides examples for both C/C++ and Rust fuzzing workflows, including standard entry point patterns like LLVMFuzzerTestOneInput and fuzz_target!. The documentation also introduces concepts such as FuzzedDataProvider for typed data extraction and interleaved fuzzing for exercising multiple operations within a single harness. By focusing on structured harness design, the skill helps developers expose deeper execution paths and improve the effectiveness of fuzz testing.

This skill is intended for security engineers, software developers, QA engineers, and researchers working with fuzzing infrastructure or vulnerability discovery. It is especially useful for teams building new fuzz targets, improving existing fuzz campaigns, or testing parsers, validators, protocol handlers, and serialization systems. The guidance is applicable to projects that require robust input handling and reproducible testing behavior across different languages and fuzzing environments.