How LLM-based penetration testing harnesses work: LLM plans, harness executes
utku1337 · reddit · 2026-10-25
utkusen's Substack piece breaks down the architecture of LLM-powered pentest tools: the user states a goal, the LLM emits commands (cat credentials.txt, curl example.com), a harness executes them in a controlled environment, and outputs — API keys, HTTP responses — are fed back for the next decision. A clear primer on the plan-execute-feedback loop behind automated offensive security agents.
More from coding & agent
- Claude Code sends 16,581 tokens per step yet scores equal a 100-line mini-harness — bendee983 · 2026-10-25
- Harrison Chase: Multiplayer agents are way harder than single-player — hwchase17 · 2026-10-25
- I measured $20 coding plans: ~33 tasks per 5h on Sonnet, only ~6 on Opus — Sufficient-Storage87 · 2026-10-25
- SwarmOS turns GitHub Issues into contracts that coding agents claim and ship as draft PRs — sn2006gy · 2026-10-25
- Adaptive Skill Granularity Framework Routes Coding Agent Skills at Runtime, Hits 100% Success — dair_ai · 2026-10-25
- Study: LLMs rewriting their own ARC-AGI-3 harnesss gain just 2.8 points, not significant — omarsar0 · 2026-10-25