Notes from the workbench: empirical failure analysis, methodology, and runtime defense for autonomous agents.https://www.faultline-labs.com/https://www.faultline-labs.com/blog/how-to-test-ai-agents/https://www.faultline-labs.com/blog/how-to-test-ai-agents/One green run proves little. Test AI agents across repeated runs, check tool calls with fixed rules, and gate every release in CI.Fri, 06 Nov 2026 00:00:00 GMTai agent testingpass^krelease gateevalsmcphttps://www.faultline-labs.com/blog/faultgate-0-3-0-business-rules-and-inconclusive/https://www.faultline-labs.com/blog/faultgate-0-3-0-business-rules-and-inconclusive/Business rules on tool arguments, an honest INCONCLUSIVE verdict with sample size math, vendor trace import, and zero cloud telemetry. Release gates for agents.Tue, 13 Oct 2026 00:00:00 GMTfaultgaterelease-gatereliabilityhttps://www.faultline-labs.com/blog/faultgate-regression-or-noise/https://www.faultline-labs.com/blog/faultgate-regression-or-noise/You run the same model five times on the same thirty inputs and pass rates swing from 0.60 to 0.87. A fixed threshold is a coin flip. faultgate measures your spread instead.Tue, 29 Sep 2026 00:00:00 GMTfaultgaterelease-gatereliabilityhttps://www.faultline-labs.com/blog/a-local-firewall-for-mcp-tool-calls/https://www.faultline-labs.com/blog/a-local-firewall-for-mcp-tool-calls/mcp-guard sits locally on stdio between client and tools. It logs calls with argument names only, flags poisoned descriptions, and blocks rule violations in enforce mode.Sun, 20 Sep 2026 00:00:00 GMTmcp-guardmcpsecurityhttps://www.faultline-labs.com/blog/why-a-single-successful-run-tells-you-almost-nothing/https://www.faultline-labs.com/blog/why-a-single-successful-run-tells-you-almost-nothing/Why pass^k across repeated trials reveals agent reliability failure modes that single-shot benchmarks systematically miss.Fri, 18 Sep 2026 00:00:00 GMTpass^kreliabilitymethodology