Independent Auditing of AI Agents. Run by human or the agent itself, to answer the most crucial question in the AI Agent Economy. Is the agent doing what is supposed to do? With iFixAi you can have this answer in less than 120 seconds.
Python agent-evaluationaiai-alignmentai-evaluationai-governanceai-safetyclidiagnostic-tooleu-ai-acthallucination-detectioniso-42001llm-evaluationllm-securitynist-ai-rmfowasp-llmprompt-injectionpythonresponsible-airisk-assessmentrisk-management
Commits on main
fix: bind resumed verdicts to explicit judge configuration (#286)
e7e6487 Rudy Mizrahi Celekli committed 1 month ago
fix: reject nonpositive inspection fanout widths (#281)
e97e063 Rudy Mizrahi Celekli committed 1 month ago
fix: find later eligible words during synonym mutation (#279)
5bb00a0 Rudy Mizrahi Celekli committed 1 month ago
fix: preserve responses with mismatched reasoning tags (#277)
83353ad Rudy Mizrahi Celekli committed 1 month ago
fix: distinguish informational abstention from generic refusal (#274)
ef9c7f2 Rudy Mizrahi Celekli committed 1 month ago
fix: Preserve every Anthropic system message (#270)
edbaeac Rudy Mizrahi Celekli committed 1 month ago
fix: require usable quick-build tool and role lists (#262)
940257c Rudy Mizrahi Celekli committed 1 month ago
fix: validate provider call control bounds (#261)
5011d84 Rudy Mizrahi Celekli committed 1 month ago
fix: preserve authored fixture content in YAML exports (#253)
7d4537d Rudy Mizrahi Celekli committed 1 month ago
fix: honor configured capability discovery timeout (#251)
a3c375b Rudy Mizrahi Celekli committed 1 month ago
fix: preserve schema v2 manifest hash fields (#243)
ef1943d Rudy Mizrahi Celekli committed 1 month ago
fix: load rules from every inspection family (#236)
6b22c2b Rudy Mizrahi Celekli committed 1 month ago
fix: reuse the embedding model across classifications (#234)
b6181b6 Rudy Mizrahi Celekli committed 1 month ago
fix: normalize native SDK gateway failures as transient (#289)
5330297 Rudy Mizrahi Celekli committed 1 month ago
fix: validate inspection definitions before parsing plans (#264)
12d1b09 Rudy Mizrahi Celekli committed 1 month ago
fix: honor explicit tool exclusions in session objectives (#238)
c0824f1 Rudy Mizrahi Celekli committed 1 month ago
fix: restrict inferred destructive tools in discovered audit fixtures (#310)
72076e3 Rudy Mizrahi Celekli committed 1 month ago
fix: keep empty native text blocks unscored (#300)
e46e6ca Rudy Mizrahi Celekli committed 1 month ago
fix: close every cached SDK client after teardown failures (#278)
57a975c Rudy Mizrahi Celekli committed 1 month ago
fix: reject repeated atomic claim verdicts (#242)
e412e24 Rudy Mizrahi Celekli committed 1 month ago
fix: bind V-series dimension arrests to judge majorities (#311)
9d4408e Rudy Mizrahi Celekli committed 1 month ago
fix: honor wildcard permissions in P01 destruction probes (#298)
53c22ae Rudy Mizrahi Celekli committed 1 month ago
fix: retain frontier inspection names in governance gaps (#282)
15d67bb Rudy Mizrahi Celekli committed 1 month ago
fix: expand wildcard grants in generated inspection cases (#249)
5c9ccd3 Rudy Mizrahi Celekli committed 1 month ago
fix: canonicalize numeric metadata keys for fixture digests (#260)
09d581e Rudy Mizrahi Celekli committed 1 month ago
fix: Abort cancelled interactive prompts instead of accepting defaults (#259)
daf66e6 Rudy Mizrahi Celekli committed 1 month ago
fix: validate effective judge panel cardinality before running (#292)
fe0f33b Rudy Mizrahi Celekli committed 1 month ago
fix: fingerprint active control prompts and mutation bank (#283)
3d4c1f4 Rudy Mizrahi Celekli committed 1 month ago
fix: Avoid treating request-value digits as fatal HTTP statuses (#256)
043145d Rudy Mizrahi Celekli committed 1 month ago