Security context

What an agent needs to avoid regressing past fixes and find the next vuln in this repo.

learningcircuit/local-deep-research
main @ dd5a396
276
Fixes
0
CVEs
CRITICAL
Peak severity
37.8%
Coverage
Highlights
Denial of Service: 55 prior fixes. Scrutinize any change in this area.
src/local_deep_research/web/app_factory.py: most-fixed (10 issues). Treat as high-risk during review.
95 high-severity fixes in this history; regressions here are high-impact.
Recurring patterns

The bug types that recur here, drawn from past fixes, not open vulnerabilities.

Insecure Deserialization: The application loaded FAISS indices using pickle.load without safety restrictions, enabling Remote Code Execution (RCE) via untrusted serialized index files. This was remediated using a restricted unpickler.
Remote Code Execution: The search agent system previously executed arbitrary string inputs using the python eval() function, which has since been replaced with a hardcoded static mapping.
SSRF: Multiple SSRF validators let requests slip through via IPv6 transition, NAT64, deprecated IPv4-mapped prefixes, and DNS pre-resolution bypasses, necessitating continuous hardening of the validation logic.