trl

Raises and scores a software Technology Readiness Level (TRL 1–9) in a loop. Each cycle collects facts-only evidence, has three low-context assessor agents score every in-scope area (core logic, data, interfaces, UI/UX, accessibility, performance, security, infra, observability, reliability, docs) without seeing the target or earlier verdicts, then writes the next level's exit criteria as failing evidence checks and builds until they pass. Works on a whole app, one feature, or a narrow improvement. Levels 7–9 are reachable before launch with production-equivalent evidence, labeled by provenance. Optional --target N (default 8), --cycles N (default 5), --assess-only, --circuit. Apply when the user invokes trl or /trl, asks for a readiness level, or wants something made ready for real users. Use opinion for open-ended takes. Use generating-app-verify to map app flows. Use proving-change-safety for one diff.