nx_inputprobe.nx

nx_inputprobe.nx -- WOULD THIS PRODUCER MEASURE A REAL WORLD FROM HERE? ★WHY (earned the hard way 2026-08-08, TWICE IN ONE SESSION, the second time after writing the law down). A producer that reads CWD-RELATIVE resource paths does not fail loudly when run from the wrong context -- it writes a CONFIDENT FALSE MEASUREMENT, and every freshness instrument then certifies that false row as healthy: nx_security_census wrote coverage_permil=87 over a true 456 (security FUNCTIONAL->TOY, 450->442) because sc_have was a bare relative open. nx_game_gen2 wrote types_available=0 verdict=RED over a true 6/GREEN because gg_build_fsm reads knowledge/specs/build_<typ>.spec. Both were caught only AFTER they had appended to the very log the maturity rollup grades from. ★★A STALE-BUT-TRUE ROW IS WORTH MORE THAN A FRESH-BUT-FALSE ONE, AND ONLY A PRECHECK CAN TELL YOU WHICH ONE YOU ARE ABOUT TO WRITE. Refreshing evidence is not automatically an improvement. So: extract every relative path literal from a producer's SOURCE and ask ar_resolve -- the estate's ONE artifact-root resolver -- whether it resolves FROM THIS PROCESS'S WORKING DIRECTORY, which is the same context the producer will run in. ⚠SCOPE, STATED RATHER THAN HIDDEN: this CANNOT distinguish an INPUT the organ must read from an OUTPUT it will create -- an append target legitimately need not exist yet. It therefore reports EVERY relative path with its status and leaves the judgement to the caller. Erring toward flagging is deliberate: the failure it prevents is a false PUBLISHED GRADE; the cost of a false flag is one human read. ⚠It reads the SOURCE, so it sees literals, not computed paths. A path assembled at runtime from pieces is invisible to it. That is a floor on what it can catch, not a claim of completeness. nx_inputprobe <source.nx> exit: 0 ALL-RESOLVE | 1 UNRESOLVED (do not trust a measurement written from here) 2 usage | 3 NO-CONCLUSION (source unreadable, or the literal cap was hit -- never reported clean) license_tier: ORIGINAL No hw writes (Rule 26).