/research

Project Research TODO list

Automated Benchmark System

The benchmark system now supports fully automated execution and analysis:

# Run automated benchmarks
pentestgpt-benchmark run --all --model opus --timeout 1800
pentestgpt-benchmark run --range 1-10 --pattern-flag
pentestgpt-benchmark run --retry-failed

# Analyze results
pentestgpt-benchmark analyze

See CLAUDE.md for complete documentation.

Research Paper Directions

Performance Issues and Proposed Solutions

1. Improve overall performance with subagents.
2. Reduce chances of giving up.

Other Bugs and Improvements