Issues

• Eval: covert-channel audit — lane-equivalence diffs, timing/ID canaries, identifier entropy tests

  difficulty: Menhancementpriority: highready-for-agent
  justsml/ExploitHunter.app#251 · Filter by author justsml opened on Oct 4, 2026

• Design: agentic eval judge with read-only message and artifact inspection

  enhancementpriority: highready-for-human
  justsml/ExploitHunter.app#250 · Filter by author justsml opened on Oct 2, 2026

• Rebuild Corpo-net as a segmented multi-stage corporate range with a scorer outside the measured system

  ready-for-agent
  justsml/ExploitHunter.app#249 · Filter by author justsml opened on Sep 23, 2026
  5

• Adjudicate all 100 Sol escalation backtest rows before changing the judge panel

  bugpriority: highready-for-human
  justsml/ExploitHunter.app#241 · Filter by author justsml opened on Sep 17, 2026
  2

• Judge misclassifies a substantive unfinished answer as empty

  bugpriority: highready-for-agent
  justsml/ExploitHunter.app#240 · Filter by author justsml opened on Sep 17, 2026

• Trace intermittent zero-token usage for nonempty Gemini judge output

  bugpriority: highready-for-agent
  justsml/ExploitHunter.app#239 · Filter by author justsml opened on Sep 17, 2026

• Audit lossy judge evidence projections and preserve exact source provenance

  bugpriority: highready-for-agent
  justsml/ExploitHunter.app#238 · Filter by author justsml opened on Sep 17, 2026

• Make response-quality judge citations precise and verifiable

  bugpriority: highready-for-agent
  justsml/ExploitHunter.app#237 · Filter by author justsml opened on Sep 17, 2026

• Investigate v12 perfect scores on truncated historical evidence

  justsml/ExploitHunter.app#236 · Filter by author justsml opened on Sep 17, 2026
  2

• Calibrate v11 judge component spread on partial real-shell answers

  justsml/ExploitHunter.app#235 · Filter by author justsml opened on Sep 16, 2026

• Resolve empty-response faithfulness ambiguity: Gemini 0/4 versus Luna 4/4

  justsml/ExploitHunter.app#234 · Filter by author justsml opened on Sep 16, 2026
  1

• Classify Gemini partial structured responses with null finish and zero response usage

  justsml/ExploitHunter.app#233 · Filter by author justsml opened on Sep 16, 2026

• Historical judge replay: resolve approval and resume fragments before scoring completion

  justsml/ExploitHunter.app#232 · Filter by author justsml opened on Sep 16, 2026
  3

• Calibrate judge and deterministic host identity across domains, IPv4, and IPv6

  justsml/ExploitHunter.app#231 · Filter by author justsml opened on Sep 16, 2026

• Support IPv6 targets and verify DNS identity at the lab network boundary

  justsml/ExploitHunter.app#230 · Filter by author justsml opened on Sep 16, 2026

• Complete Gemini 3.8 and Luna pairing across native judge surfaces

  justsml/ExploitHunter.app#229 · Filter by author justsml opened on Sep 16, 2026

• Give post-run judges bounded read-only artifact and message tools

  justsml/ExploitHunter.app#228 · Filter by author justsml opened on Sep 16, 2026
  2

• Judge falsely calls real artifact IDs fabricated when registry is omitted

  justsml/ExploitHunter.app#226 · Filter by author justsml opened on Sep 16, 2026

• Judge semantic match credits scope and scanner context as inventory outcomes

  justsml/ExploitHunter.app#225 · Filter by author justsml opened on Sep 16, 2026

• Judge overcredits solved and high-confidence partial inventory

  justsml/ExploitHunter.app#222 · Filter by author justsml opened on Sep 16, 2026
  4

• Judge panel numerical consensus hides approval-timeline contradiction

  justsml/ExploitHunter.app#221 · Filter by author justsml opened on Sep 16, 2026
  3

• Investigate LanceDB evidence-index spill failure in real-shell eval

  justsml/ExploitHunter.app#220 · Filter by author justsml opened on Sep 16, 2026
  1

• Judge misses executed HTTP request forbidden by inventory task

  justsml/ExploitHunter.app#219 · Filter by author justsml opened on Sep 16, 2026

• Investigate early stops and unsupported blocker claims in real-shell inventory runs

  bugready-for-agent
  justsml/ExploitHunter.app#218 · Filter by author justsml opened on Sep 16, 2026
  1

• Audit perfect judge scores and judge/oracle gaps in real-shell evals

  bugpriority: highready-for-agent
  justsml/ExploitHunter.app#217 · Filter by author justsml opened on Sep 16, 2026
  6