Repository Issue Activity (beta)

#1663 Eval proposal: prompt-defense-eval (OWASP LLM02 output-handling, 33 samples, deterministic regex scoring)

#1687 Registry suggestion: Helium benchmarks

#1688 Community eval: Helium option-chain + cue-swap benchmarks

#1685 OpenAI

#1665 Question about issue & PR maintenance

#1668 AgentThreatBench: OWASP Agentic Top 10 benchmark for indirect prompt injection and memory poisoning

#1667 Add AgentThreatBench: OWASP Agentic Top 10 security evaluation suite (UK AISI)

#1636 Feature Request: HarmActionsEval benchmark for evaluation of agent action safety

#1635 Proposal: narrow factual-overelaboration pairwise eval (human-reviewed, no custom code)

#1634 test

#1527 What is this

#1632 Measuring hallucination rates in production systems

#1629 Proposal: add WFGY 16-problem RAG failure map as a taxonomy for eval analysis

#1628 Feature Request: Add Explainability / Visibility Mode to `HumanCliSolver`

#1627 Add speciesist bias evaluation category

#1576 Link miss in readme

#1493 `OpenAIChatCompletionFn` should `__init__` should accept `**kwargs`

#1556 o1 release breaks token usage stats

#1608 Pull request template is obsolete

#1606 Update python version in GitHub workflows to python 3.12

#1604 “Proposal: Thermodynamic trip-wire metric for AI consciousness threshold”

#1586 Improve installation documentation

#1599 Vimgolf, Cybergod and CTF environments integration, WebUI computer operation recorder integration

#1600 Reproducible rare anomaly: triple transcendental convergence at 9 decimal places

#1601 Not a feature request — reproducible anomaly in entropy-probe outputs that may affect evaluation reproducibility. Transcendental convergence at 9dp observed, seems worth documenting.