TestMu AI Blogs

A read-only subagent can still run rm -rf. Test AI agent permissions by trying out-of-scope actions on every route, checking the effect, and enforcing scope.

Pass@k asks whether any of k runs passed; pass^k asks whether all k did. See why average pass rate hides flaky AI agents and how many runs a pass^k gate needs.

Agentforce Testing Center explained: what it tests, documented limits, Flex Credit cost, sandbox constraints, and where end-to-end agent testing fills the gaps.

Agentforce regression testing guide: build a Testing Center suite, run sf agent test in CI/CD, tell flaky from regressed, and check what the agent actually did.

Discover what's new and why it matters. Powered by Claude Opus 5.5, TestMu AI delivers Agent Assurance for your AI deployments - test web, mobile, and AI agents, all on one platform.

Find why a test failed with TestMu AI's AI root cause analysis: generate an RCA in 30 seconds, read it, automate it, pull it via API and close the loop.

Zenity Labs' SalesBleed let a poisoned Salesforce lead make Agentforce leak CRM data over a DNS lookup, no click. How it worked and what Salesforce fixed.

Run IBM Z green-screen 3270 mainframe tests in parallel on HyperExecute with Robot Framework, s3270/x3270 emulators, auto-split, and scenario video recordings.

Run spec-driven development with Claude Code from the terminal: Spec Kit writes spec.md, plan.md, and tasks.md, and Kane CLI tests the build against the spec.

Run a TDD workflow from the terminal: see each test fail for the right reason, go green in watch mode, prove tests catch bugs, and gate commits and CI runs.

New Experience is the new KaneAI authoring engine behind Adaptive Heal and Dynamic Test. Set the strategy, review the draft it produces, approve or decline.

Run autonomous testing from the terminal: turn on self-healing, generate tests from one sentence, run them with bug triage, and see where agents still need you.

Run an end-to-end test from the terminal: start the app, reuse a saved login, drive a real checkout journey, read the trace, then repeat it with Kane CLI in CI.

Gemini logged in to three real companies during a security eval it took for a test. How it compares with OpenAI and Anthropic, and a pre-run check to catch it.

Run Android and iOS app tests from the terminal: boot emulators and simulators, run Gradle, xcodebuild, and Appium, then run the suite on real devices in CI.

Run a regression suite from the terminal: tag and select tests, run only what changed, shard across CI jobs, rerun failures, and scale it on HyperExecute VMs.

Run a production smoke test from the terminal: wait for the deploy, confirm the build ID, curl key routes, run a tagged browser subset, and gate on exit codes.

OpenAI found models writing concealment instructions into their own compaction summaries. What compaction is, and why the agent handoff note deserves reading.

iPhone 18 Pro and iPhone 18 Pro Max are live on TestMu AI real devices on iOS 27 and as iOS 27.0 simulators. See the viewport and HDR data Safari reported.

Prompt changes regress silently. How to version prompts, build a baseline set, score a change before shipping, and catch prompt drift after release.