https://joe-b-security.github.io/tags/benchmarks/Recent content in Benchmarks on Joe Bollen SecurityHugo -- gohugo.ioenYes, it's a real three-body problem simulationMon, 13 Apr 2026 00:00:00 +0000https://joe-b-security.github.io/posts/2026-04-13-improving-coding-agent-harness-part3/Mon, 13 Apr 2026 00:00:00 +0000https://joe-b-security.github.io/posts/2026-04-13-improving-coding-agent-harness-part3/<p>In <a href="https://joe-b-security.github.io/posts/2026-04-07-improving-coding-agent-harness-part1/">Part 1</a>, I added tree-sitter tools for structural code reading. In <a href="https://joe-b-security.github.io/posts/2026-04-07-improving-coding-agent-harness-part1-5/">Part 1.5</a>, I locked those tools behind a secure factory. In <a href="https://joe-b-security.github.io/posts/2026-04-09-improving-coding-agent-harness-part2/">Part 2</a>, I added an OODA loop with a rule engine and verify phase. In <a href="https://joe-b-security.github.io/posts/2026-04-10-improving-coding-agent-harness-part2-5/">Part 2.5</a>, I added a RAG layer over OWASP guidance for secure code generation.</p>