Skip to content

Basilisk - v1.1.0

Latest
Latest

Choose a tag to compare

View all tags
@kalpmodi kalpmodi released this 25 Apr 08:17
· 4 commits to dev since this release
v1.1.0
ac142fe

Basilisk v1.1.0

Architecture Overhaul: Thin Router + Technique Library

The biggest internal upgrade since launch. Every large skill has been split into a thin router and a technique library loaded on demand.

Token reduction per session: 75-80%

Skill Before After
zerodayhunt 1843 lines 179-line router + 20 tech files
exploit 1200 lines 234-line router + 20 tech files
recon 890 lines 119-line router + 14 tech files
redteam new 207-line router + 14 tech files

New: _shared/ Canonical Library

  • _shared/phase0.sh - unified session state, intel relay, memory read for all 15 skills
  • _shared/signals.sh - atomic POSIX-append signal emission, race-proof across concurrent skill forks

All skills now source these instead of duplicated Phase 0 bash blocks. Intel relay is consistent across every skill - recon findings flow automatically into exploit, cloud-audit, zerodayhunt.

New Skills

  • /redteam - Full red team: initial access, lateral movement, ADCS ESC1-8, C2, persistence, cloud APT (Azure Device Code, Pass-the-PRT), exfil, OPSEC
  • /compact - Compresses session.json mid-engagement when context grows large

68 Technique Files Added

New tech/ directories under exploit, recon, redteam, zerodayhunt covering:

  • JWT RS256->HS256 algorithm confusion + SAML XSW
  • Prototype pollution -> EJS/Pug RCE gadgets (USENIX 2023)
  • mXSS DOMPurify bypass CVE-2024-47875
  • ECDSA nonce reuse + CVE-2022-21449 Psychic Signatures
  • ImageMagick CVE-2022-44268 PNG tEXt arbitrary file read
  • HTTP Desync CL.0 + H2.CL (Black Hat 2022)
  • CI/CD pull_request_target fork RCE
  • XS-Leaks cache probing + timing oracles
  • 11 confirmed attack chain blueprints (Chain A-K)
  • PDF generator SSRF -> AWS IAM credentials
  • Gopher protocol SSRF -> Redis RCE
  • Nginx alias off-by-slash traversal
  • CSS injection CSRF token exfiltration

Hypothesis Engine Improvements

  • Bayesian posterior update: P(H|e) = P(H) × ∏ P(eᵢ|H) / P(eᵢ) in log-space
  • Signal likelihood table with per-hypothesis calibration
  • Manifest reprioritization: SHOULD -> MUST when P(H) > 0.45
  • Doom loop ATW write-back persisted to memory across engagements

Full Changelog

v1.0.0...v1.1.0

Assets 2
Loading