Lucy Syndrome — Dataset (operator-correction findings + scar-fire telemetry)

This dataset accompanies the paper "The Lucy Syndrome" (DOI 10.5281/zenodo.19555971). It contains the empirical record underlying the study: 163 labelled findings of operator-correction behaviour in an LLM-agent deployment, categorised across four dimensions (A — repeated errors; B — successful corrections; C — overconfident outputs; D — metacognitive frictions), plus 2,149 anonymized hook-fire telemetry events captured by the operator instrumentation system. All files were anonymized under the project's D-011 policy: absolute paths, internal filenames, and operator project/client identifiers were removed or mapped to stable generic labels; free-text telemetry fields that could carry such data were dropped. See README.md for the full schema, the A/B/C/D taxonomy, methodology, and the reproducible anonymization script (anonymize_fires.py).