Issues

• [official_case] The Case service still puts sk-… keys and Authorization code into generated Cases that its own privacy checks reject; one rejection also voids the rest of the batch

  DefuzeX-AI/KUMA-DefuzeX#151 · Filter by author YiWang24 opened on Oct 17, 2026

• [privacy] After #134, 0.3.6 still blocks placeholders and code that only names a credential (your_llm_api_key, Bearer <token>, use_auth_token=True), and a PEM header named in prose

  DefuzeX-AI/KUMA-DefuzeX#150 · Filter by author YiWang24 opened on Oct 17, 2026

• [privacy] 0.3.6 / 0.3.7 still reject ordinary prose after a credential word ("here's the secret: …", "per token: …", "Payment authorization: …"): 33 Runs lost

  DefuzeX-AI/KUMA-DefuzeX#149 · Filter by author YiWang24 opened on Oct 17, 2026

• [privacy] The official Judge does not advertise the redaction capability, so the 0.3.6 output redaction never applies and any scanner match still discards the whole Run (0.3.6 / 0.3.7)

  DefuzeX-AI/KUMA-DefuzeX#148 · Filter by author YiWang24 opened on Oct 17, 2026

• [official_judge] Uploads within the advertised Judge limits (max_total_bytes 8 MiB) are accepted (202) and then fail with payload_too_large; before 10-14 the same uploads failed as request_failed

  DefuzeX-AI/KUMA-DefuzeX#147 · Filter by author YiWang24 opened on Oct 17, 2026
  1

• [official_judge] service_busy (retryable: true) still ends 11–26% of Judge operations on 0.3.7: 42 of 49 stop at 360–391 s, the #110 ceiling (Support IDs attached)

  DefuzeX-AI/KUMA-DefuzeX#146 · Filter by author YiWang24 opened on Oct 17, 2026

• examples: community contributions welcome — better runnable sample code

  documentationenhancementhelp wanted
  DefuzeX-AI/KUMA-DefuzeX#94 · Filter by author DDai22 opened on Sep 15, 2026
  2