DATA OBFUSCATOR — RUN SUMMARY
=============================
Generated: 2026-10-13 06:54:28 UTC
Program:   data-obfuscator 0.0.0.23 (released 2026-10-13)
Command:   run

INPUT
  File:            pk_study.csv
  Format:          CSV (comma)
  Rows:            180
  Columns (in):    22
  Subjects (ID):   20
  Bytes:           27894

OUTPUT
  File:            obfuscated.csv
  Rows:            180
  Columns (out):   22
  Subjects (ID):   20
  SHA-256:         466efb82b82d9dc7d9846c990eeda57de74db9efd452f7d131b91083bbd7e55c

OBFUSCATION
  Level:           2 (clinical labels)
  Seed:            2026
  ID style:        sequential
  Columns recoded: ID, USUBJID, STUDYID, SITEID, ARM, SEX, RACE, DATE, INVESTIGATOR, COUNTRY
  Columns dropped: (none)
  Columns kept:    ID, USUBJID, STUDYID, SITEID, ARM, TIME, TAD, AMT, DV, EVID, MDV, CMT, WT, AGE, SEX, RACE, DATE, INVESTIGATOR, COMMENT, COUNTRY, EMAIL, SUBJINIT
  Date shift:      50.8 days (one decimal value for the whole dataset; intervals preserved; exact value in keymap.txt)
  Scaling:         off
  Keymap written:  yes — DO NOT share keymap.txt with any LLM or external party
  Per-value file:  none

REPRODUCIBILITY
  Same input file + same options + same seed + same program version gives
  a byte-identical dataset. Compare the SHA-256 above between runs to check.

INTEGRITY
  Within-ID time order preserved: yes
  Protected columns untouched:    TIME, TAD, AMT, DV, EVID, MDV, CMT
  Unique ID count unchanged:      yes

TOKEN COUNTS
  See tokens.txt for estimated token counts (GPT, Claude, Gemini, Llama,
  Qwen, Mistral) of the input file, the output dataset, and every other
  file in the output directory except the log.

LLM PROMPTS
  See llm.txt for ready-to-paste prompts (analysis, NONMEM control
  stream, QC) tailored to the masked dataset.

STATUS: SUCCESS