orchestrate-dataset-coupling-auditorlisted

# Orchestrate: Dataset Coupling Auditor **Evidence tier: first-hand build (August 2026).** Grounded in a completed Orchestrate submission that was audited to destruction — 48 logged defects, 9 measured-and-rejected optimisations, 17 certification scripts. Every number below was measured on that system. Nothing here claims access to HackerRank's internal scoring. ## The rule **Assume the hidden data differs from your sample in every way the specification does not explicitly fix. Then go looking.** ## The taxonomy — and the test for each | Coupling | Test | |---|---| | ID **format** | rename every id to a different scheme; do decisions survive? | | Timestamp format | rewrite dates day-first; does ordering survive? | | Row order | shuffle every context file; does output survive? | | Directory layout | rename the media folder; does resolution survive? | | Filename ↔ content | verify by magic bytes, never extension | | Exact wording | which lexicon terms fire on exactly **one** sample row? | ## What this caught in a real build — three real couplings, all found late **1. ID shape.** Mention detection harvested tokens matching a regex for the sample's id convention (`@u_<digits>`), then compared them to the recipient. The comparison was principled; the *harvest pattern* was not. Renaming ids broke **4 of 110 decisions** — two rules went silently dead. Fixed by searching for `@` + *the actual recipient id*, assuming no format at all. **2. Timestamp format.** History was sorte