What problem does it solve? Moving a debugging runtime from demo smoke tests into a real customer app exposes gaps that self-tests never find: stale targets, unverified business state, leaked private data in public reports, and unproven regression claims. This Skill guides an AI agent through disciplined real-project regression testing of TritonKit so every finding is backed by machine-readable evidence and every product gap becomes a doc update, fix, or redacted GitHub issue. ## Core Features & Use Cases - Isolated real-app validation: Keeps external app repo changes separate from TritonKit commits, prefers the released Homebrew triton CLI, and verifies binary identity with triton version --json before any flow. - Triton-first command workflow: Preserves doctor, status, capabilities, and schema JSON facts before any fallback to raw xcrun, hdc, or xcodebuild, and treats schema contract anomalies as reportable TritonKit bugs. - Evidence-gated completion: Uses triton evidence capture, screenshots, wait/verify, and replay plans so a host action acknowledgement is never mistaken for a business verdict. - Redaction and issue filing: Enforces redaction of bundle IDs, private paths, accounts, and unredacted evidence before anything is published, routing gaps into the tritonkit-dev-feedback flow. - Use Case: An agent validating TritonKit in a customer's iOS app boots the simulator through triton sim, connects the DEBUG embedded runtime, runs an observe-act-verify flow, captures a .tritonevidence bundle, redacts it, and files a sanitized issue for a discovered capability gap. ## Quick Start Ask the agent to run a real-project regression of TritonKit against your iOS or Harmony app, starting with triton doctor and capabilities checks and ending with redacted evidence for every finding.
