graph-guard is actively maintained. Security fixes target the latest released version on the main branch.

Please do not open a public issue for security vulnerabilities.

Report privately through GitHub's Report a vulnerability flow (the repository's Security → Advisories tab). I aim to acknowledge reports within 72 hours and to ship a fix or mitigation for confirmed issues as quickly as is practical.

When reporting, please include:

• a description of the issue and its impact,
• steps to reproduce (a minimal proof-of-concept if possible), and
• any suggested remediation.

graph-guard builds a typed knowledge graph over a local note vault and exports it to RDF/OWL/SHACL/SPARQL. Findings of particular interest:

• Private-data leakage past the eval harness's aggregate-only guarantee — eval/results.json and the retrieval-lift docs must never contain note ids, file paths, or query/gold text (enforced by a leak-check test).
• Grounding or structural-refuse bypass in the guards (graph_guard/guards.py).
• Unsafe deserialization when parsing untrusted RDF/Turtle, or SPARQL injection via unsanitized query terms (IRIs are serialized through rdflib's N3 form for this reason).
• Supply-chain risk in CI — this repository pins its GitHub Actions to commit SHAs and runs CodeQL + Dependabot, and pulls its guarded-rag dependency from a Git source (github.com/Jott2121/rag-guard).

Thanks for helping keep it solid.