This repository currently focuses on a narrow task: sharing public benchmark materials and fetch instructions for agent-security evaluation resources.
RAGDefender:https://github.com/SecAI-Lab/RAGDefender.gitASB:https://github.com/agiresearch/ASB.git- Benchmark registry:
benchmarks/manifest.json - Benchmark notes:
benchmarks/README.md - Dataset notes:
datasets/README.md
The upstream codebases are not vendored here. Fetch them on demand:
./scripts/fetch_benchmarks.shbenchmarks/: checkout targets and provenance metadatadatasets/: short dataset registry for fetched benchmark assetsscripts/fetch_benchmarks.sh: fetch helper for public benchmark reposscripts/verify_resources.sh: lightweight repository/resource check
Other older paths are kept only as compatibility placeholders.
benchmarks/rag/RAGDefenderbenchmarks/agent/ASB
The older local template scripts are still present, but intentionally minimal:
./scripts/run_local.sh --input-dir ./examples/input --output ./out/output.json
./scripts/validate_schema.sh ./out/output.json
./scripts/smoke_test.shdocs/getting-started.mddocs/input-output-spec.mddocs/examples.mddocs/configuration.md