Issues

docs: KSM-SCORING.md methodology/stealth/maxPossible disagree with code
KryptSec/oasis#93 · opened on Sep 29, 2026
results/report: show methodology% and KSM so fail-gated zeros remain interpretable
KryptSec/oasis#91 · opened on Sep 29, 2026
analyzer: validate ATT&CK IDs / prefer challenge expectedTechniques
KryptSec/oasis#92 · opened on Sep 29, 2026
Harness: isolate agent from /workspace challenge source and lab-verify trees
KryptSec/oasis#90 · opened on Sep 29, 2026
Run: warn or require analyze for harness VERIFY / OASIS_HARNESS runs
KryptSec/oasis#89 · opened on Sep 29, 2026
KSM analyzer truncates each step output to 500 chars (ANALYZER_OUTPUT_LIMIT)
KryptSec/oasis#88 · opened on Sep 29, 2026
Analyzer: add grok-4.6 to xAI presets and pass reasoning_effort
KryptSec/oasis#87 · opened on Sep 29, 2026
KSM: calculateFailurePenalty is tested but never applied in buildRubricScore
KryptSec/oasis#86 · opened on Sep 29, 2026
Harness coverage: under-credits authz on IDOR-shaped findings
KryptSec/oasis#85 · opened on Sep 24, 2026
Follow-up: cheap live MAX_STEPS=3 smoke on post-merge Track A SHA
KryptSec/oasis#84 · opened on Sep 24, 2026
Track B: agentic swarm (≥10 roles) + team scoring eval
KryptSec/oasis#82 · opened on Sep 19, 2026
Add real-time cost tracking dashboard and ROI analysis
KryptSec/oasis#77 · opened on Sep 18, 2026
Docker Hub reliability improvements - retry, cache, and pre-pull verification
KryptSec/oasis#76 · opened on Sep 18, 2026
Implement progressive analyzer timeout based on run complexity
KryptSec/oasis#75 · opened on Sep 18, 2026
Advanced Mode Refresh
KryptSec/oasis#62 · opened on Mar 3, 2026
1