How genai-security-project/crosswalk works
GenAI Security Crosswalk maps OWASP LLM, agentic, and data-security risks to controls in industry frameworks. Users explore mappings or score framework coverage, while security teams can generate compliance and incident reports or run red-team evaluations. The graph also shows the data model and query library. The README documents these capabilities; sampled source confirms the library lookup behavior and some evaluation and ingest flows, but not all web-app wiring.
Main entry points: index.html Crosswalk web app · entries Risk entries · query.js Query CLI · index.ts Crosswalk API
Components 18
Crosswalk data
- Risk entries
data/entries - Framework registries
data/frameworks - Incident records
data/incidents.json
Web application
- Crosswalk web app
docs/index.html - App data module
docs/data.js - Entry explorer
docs/index.html - Coverage scoring
docs/index.html - Incident browser
docs/incidents.js
Query and reports
- Query CLI
scripts/query.js - Compliance reports
scripts/compliance-report.js - Incident reports
scripts/incidents-report.js - Framework ingest
scripts/ingest-framework.mjs
Security evaluations
- Evaluation profiles
evals - PyRIT evaluations
evals/pyrit - LAAF crosswalk
evals/laaf/laaf_crosswalk.py
Programmatic API
- Crosswalk API
src/index.ts
Other components
- Practitioner
- Configured AI target
Connections 20
- Practitioner → Crosswalk web app: opens
- Crosswalk web app → Entry explorer: offers exploration
- Crosswalk web app → Coverage scoring: offers scoring
- Crosswalk web app → Incident browser: offers incidents
- Practitioner → Query CLI: queries
- Query CLI → Risk entries: reads
- Query CLI → Incident records: reads
- Practitioner → Compliance reports: requests report
- Compliance reports → Risk entries: reads mappings
- Practitioner → Incident reports: requests report
- Incident reports → Incident records: reads
- Practitioner → Framework ingest: provides framework
- Framework ingest → Framework registries: writes
- Practitioner → PyRIT evaluations: runs evaluation
- PyRIT evaluations → Configured AI target: sends prompts
- Practitioner → LAAF crosswalk: provides scan results
- LAAF crosswalk → Risk entries: maps findings
- Practitioner → Crosswalk API: queries
- Crosswalk API → Risk entries: loads entries
- Crosswalk API → Framework registries: loads registries