Skip to main content

Videos & Presentations

Short narrated walkthroughs of setting up OSS-CRS and running a CRS end to end, plus slides from public talks.

The videos follow one workflow: set up OSS-CRS, run the LibFuzzer bug-finding CRS against a target, inspect the seeds and proofs of vulnerability (POVs) it produces, and then pass those POVs to a Claude Code bug-fixing CRS that writes a patch. Watch them in order, or jump to the step you need.

Training videos​

#VideoLengthWhat it covers
1Set up OSS-CRS3:15oss-crs setup: routing LLM providers through a proxy (provider selection, API key and base URL env vars) and the cgroup v2 delegation step for CPU and memory isolation (Linux x86 only).
2Set variables for the LibFuzzer CRS1:15Defining the compose file (CRS), the harness (HARNESS) and the target project (TARGET) for a bug-finding campaign.
3Prepare the LibFuzzer CRS0:41oss-crs prepare: building the CRS base image and pre-caching tools so later phases don't rebuild.
4Build a target for LibFuzzer1:42oss-crs build-target: compiling the target with sanitizers (ASan, UBSan, MSan) and coverage instrumentation, and linking it against the fuzzing harness.
5Run the LibFuzzer CRS2:03oss-crs run: initializing the CRS, validating the configuration, preparing the run environment and starting the fuzzer with its sidecars.
6LibFuzzer outputs: seeds and POVs2:13What seeds and proofs of vulnerability are, where they are stored under SUBMIT_DIR, and how to view them in the web UI dashboard.
7Run a patching CRS4:44Passing LibFuzzer POVs to the Claude Code bug-fixing CRS with --pov-dir, using the LiteLLM proxy, and reviewing the generated patch.

Slides​