PBFuzz: Agentic Directed Fuzzing for PoV Generation Artifact for the CCS 2026 paper. PBFuzz is a directed fuzzer that uses an LLM agent to generate property-based test inputs targeting specific bug locations. Given a bug ID and static call-graph analysis, the agent iteratively synthesizes Python input generators, executes the target under Magma canary instrumentation, and uses GDB/LLDB feedback to guide generation toward triggering the bug. This artifact contains the PBFuzz source code, Docker build scripts, and pre-built static analysis results for 9 Magma bugs (LUA001, PDF002, PHP014, PNG006, SND001, SQL012, SSL002, TIF006, XML002). Magma benchmark cloned from GitHub as part of setup.This research artifact is also actively maintained at: https://github.com/sgzeng/pbfuzz.gitPaper preprint: https://arxiv.org/abs/2512.04611