1.9.0 - Slipstream
What's Changed
- fix(prompt): align attacker/reviewer structured-output contract by @vinceAmstoutz in #41
- feat(prompt): keep uncertain findings instead of dropping them by @vinceAmstoutz in #42
- fix(llm): estimate Fable and Mythos tokens with their denser tokenizer by @vinceAmstoutz in #43
- docs(faq): add per-agent reasoning-effort / thinking guidance by @vinceAmstoutz in #45
- feat(agent): feed reviewer-rejected findings back to the attacker by @vinceAmstoutz in #44
- feat(prompt): mark firewall-covered routes in the attacker route map by @vinceAmstoutz in #47
- feat(agent): add provider-validated reviewer verdicts via record_review tool by @vinceAmstoutz in #48
- feat(prompt): add opt-in stable system prompt for provider cache reuse by @vinceAmstoutz in #46
- feat(cache): cache reviewer verdicts across runs by @vinceAmstoutz in #49
- feat(report): list vulnerabilities most-severe-first in every output format by @vinceAmstoutz in #54
- docs: rename Ollama platform option host_url to endpoint by @vinceAmstoutz in #53
- feat(llm): price prompt-cache tokens into audit cost by @vinceAmstoutz in #50
- feat(command): surface live audit-stage progress and a long-run notice by @vinceAmstoutz in #55
- feat: config-less cheaper and faster defaults by @vinceAmstoutz in #56
Full Changelog: 1.8.0...1.9.0