feat(content,pipeline): false-citations content type + Proverbs 3:13 / pseudo-Enoch seeds

[CraigBuckmaster]

Strategic context

Viral religious content consistently uses two fabrication patterns that are categorically distinct from existing proof-text-guards coverage:

1. Misquoted canonical verses — a real verse cited with words that are not in the verse. Recent example: "Proverbs 3:13: Blessed are those who seek wisdom, for they will not be deceived. Blessed are those who know the enemy, for they will escape his snares." Proverbs 3:13 actually reads "Blessed are those who find wisdom, those who gain understanding" (NIV) — the rest is fabricated.

2. Misattributed extrabiblical quotations — words attributed to 1 Enoch / Jasher / Gospel of Thomas / Dead Sea Scrolls that do not appear in any extant text. Recent example: "Enoch wrote: They will lead many astray, and the truth will be hidden from the children of men." Not present in 1 Enoch (Ge'ez, Aramaic, or Greek).

proof-text-guards.json covers real verses misapplied through reading. It does not cover fabricated wording or false attribution. The schemas differ enough that conflation would erode the proof-text-guards entries' usefulness.

This card adds a new content type — false-citations — with the schema, two seed entries, build-pipeline wiring, and a SQLite landing table.

Parent epic: #1536
Related cards: #1817, #1819, #1820, #1821 (Divine Council / cosmic-conflict content) — those cards may want to cross-reference false-citations entries

Deliverables

1. New metadata file: content/meta/false-citations.json
2. Two seed entries: misquoted Proverbs 3:13; pseudo-Enoch "children of men"
3. Schema-validator coverage in _tools/schema_validator.py
4. SQLite loader in _tools/build_sqlite_loaders.py + table DDL in _tools/build_sqlite_schema.py
5. SQLite validator check in _tools/validate_sqlite.py

All deliverables ship in the same PR.

---

Deliverable 1 — Schema

File created: content/meta/false-citations.json

Top-level type: JSON array (match difficult-passages.json and proof-text-guards.json convention).

Entry schema:

{
  "id": "string-kebab-case",                          // unique, required
  "category": "misquoted-canonical | misattributed-extrabiblical | fabricated-no-source",
  "alleged_text": "the wording as it circulates",     // required
  "claimed_source": "human-readable source string",   // e.g. "Proverbs 3:13" or "1 Enoch"
  "claimed_source_book_id": "proverbs | null",        // null for non-canonical sources
  "claimed_source_chapter": 3,                          // null if not applicable
  "claimed_source_verse": 13,                           // null if not applicable
  "status": "fabricated | misquoted | misattributed",
  "actual_text": "the real verse, or null if no source exists", // null for fabricated
  "actual_text_translation": "NIV | ESV | original",
  "actual_text_book_id": "proverbs | null",
  "actual_text_chapter": 3,
  "actual_text_verse": 13,
  "what_is_wrong": "1-3 sentence explanation",        // required
  "why_it_circulates": "1-3 sentence pastoral note",  // required, charitable framing
  "related_difficult_passage_ids": ["ethiopian-bible-true-canon"], // optional cross-refs
  "tags": ["string"]                                    // optional
}

Schema rules:

• id must be unique across the file
• category enum strict; reject any other value
• claimed_source_* fields all required to be either populated or null together (no partial population)
• actual_text_* fields all required to be either populated or null together
• If status == "fabricated", actual_text and all actual_text_* MUST be null
• If status == "misquoted", actual_text MUST be populated
• If status == "misattributed", actual_text is typically null but may be populated if the misattribution swaps a real quote between sources

Deliverable 2 — Seed entries

Entry 1 — fabricated Proverbs 3:13

{
  "id": "proverbs-3-13-seek-wisdom-deceived",
  "category": "misquoted-canonical",
  "alleged_text": "Blessed are those who seek wisdom, for they will not be deceived. Blessed are those who know the enemy, for they will escape his snares.",
  "claimed_source": "Proverbs 3:13",
  "claimed_source_book_id": "proverbs",
  "claimed_source_chapter": 3,
  "claimed_source_verse": 13,
  "status": "misquoted",
  "actual_text": "Blessed are those who find wisdom, those who gain understanding.",
  "actual_text_translation": "NIV",
  "actual_text_book_id": "proverbs",
  "actual_text_chapter": 3,
  "actual_text_verse": 13,
  "what_is_wrong": "Proverbs 3:13 commends finding wisdom and gaining understanding. It does not contain the words 'deceived,' 'enemy,' or 'snares.' The circulating wording appears to be a free composition styled to fit conspiracy-adjacent content.",
  "why_it_circulates": "The fabricated wording functions rhetorically — it lets a viral post close on a 'biblical' note that endorses the post's framing about hidden truth and enemies. Readers who have not opened a Bible to Proverbs 3 in some time will not catch the swap.",
  "related_difficult_passage_ids": [],
  "tags": ["fabricated", "viral-content", "wisdom-literature"]
}

Entry 2 — pseudo-Enoch

{
  "id": "pseudo-enoch-children-of-men",
  "category": "misattributed-extrabiblical",
  "alleged_text": "They will lead many astray, and the truth will be hidden from the children of men.",
  "claimed_source": "Book of Enoch",
  "claimed_source_book_id": null,
  "claimed_source_chapter": null,
  "claimed_source_verse": null,
  "status": "fabricated",
  "actual_text": null,
  "actual_text_translation": null,
  "actual_text_book_id": null,
  "actual_text_chapter": null,
  "actual_text_verse": null,
  "what_is_wrong": "These words do not appear in any extant text of 1 Enoch — neither in the Ge'ez full text, the Aramaic Qumran fragments (4QEn^a-g), nor the Greek fragments (Codex Panopolitanus, Chester Beatty). 1 Enoch's actual treatment of the Watchers' end (1 En. 10, 16) describes their binding and the judgment of their offspring's spirits, not their continuing rule over humanity.",
  "why_it_circulates": "The line is styled to sound like 1 Enoch — biblical cadence, 'children of men' phrasing, generic apocalyptic warning. It lets a viral post claim Enoch endorses a hidden-truth / leading-astray framing the actual text does not support. Few readers can verify against the real Ge'ez or Aramaic text.",
  "related_difficult_passage_ids": ["jude-quotes-enoch", "ethiopian-bible-true-canon"],
  "tags": ["fabricated", "viral-content", "extrabiblical", "1-enoch"]
}

Deliverable 3 — Schema validator

File modified: _tools/schema_validator.py

Add a validate_false_citations() function following the existing pattern (modelled on the closest existing validator — likely validate_difficult_passages or validate_proof_text_guards). Enforce:

• File parses as JSON array
• Each entry has required fields per the schema above
• id values are unique
• category and status enum values are valid
• The conditional-population rules: claimed_source_* fields populated-together-or-all-null; same for actual_text_*
• status == "fabricated" ⇒ actual_text is null
• status == "misquoted" ⇒ actual_text is non-null
• claimed_source_book_id, when non-null, resolves in books.json
• actual_text_book_id, when non-null, resolves in books.json
• All related_difficult_passage_ids resolve in difficult-passages.json
• All tags are non-empty strings

Add the validator to the main validation runner so it executes by default.

Deliverable 4 — SQLite loader and schema

Files modified: _tools/build_sqlite_loaders.py, _tools/build_sqlite_schema.py

Table DDL (in build_sqlite_schema.py):

CREATE TABLE IF NOT EXISTS false_citations (
  id TEXT PRIMARY KEY,
  category TEXT NOT NULL,
  alleged_text TEXT NOT NULL,
  claimed_source TEXT NOT NULL,
  claimed_source_book_id TEXT,
  claimed_source_chapter INTEGER,
  claimed_source_verse INTEGER,
  status TEXT NOT NULL,
  actual_text TEXT,
  actual_text_translation TEXT,
  actual_text_book_id TEXT,
  actual_text_chapter INTEGER,
  actual_text_verse INTEGER,
  what_is_wrong TEXT NOT NULL,
  why_it_circulates TEXT NOT NULL,
  tags_json TEXT
);

CREATE TABLE IF NOT EXISTS false_citation_related (
  false_citation_id TEXT NOT NULL REFERENCES false_citations(id),
  difficult_passage_id TEXT NOT NULL,
  PRIMARY KEY (false_citation_id, difficult_passage_id)
);

CREATE INDEX IF NOT EXISTS idx_false_citations_status ON false_citations(status);
CREATE INDEX IF NOT EXISTS idx_false_citations_book ON false_citations(claimed_source_book_id);

Loader (in build_sqlite_loaders.py):

• Add load_false_citations(conn, content_root) following the pattern of existing loaders (load_difficult_passages is the closest precedent)
• Reads content/meta/false-citations.json
• Inserts one row per entry into false_citations
• Serialises tags as JSON into tags_json
• Inserts one row per related_difficult_passage_ids element into false_citation_related
• Wire the new loader into the main pipeline in build_sqlite.py

Deliverable 5 — SQLite validator

File modified: _tools/validate_sqlite.py

Add an integrity check that:

• false_citations table exists and has the expected column set
• All rows in false_citation_related.difficult_passage_id resolve to a row in difficult_passages (assuming that table name — verify against existing loader output)
• Both seed entries are present after build

---

Acceptance criteria

• content/meta/false-citations.json exists with the two seed entries
• schema_validator.py runs clean against the file
• schema_validator.py rejects deliberately-broken test cases (verify by running locally with a temporary invalid entry — do not commit the broken case)
• build_sqlite.py runs clean; false_citations table populated with 2 rows; false_citation_related table populated with 1 row (from the pseudo-Enoch entry's cross-refs)
• validate_sqlite.py passes including the new integrity check
• No new image URLs introduced
• PR description includes a one-line note on the schema decision (why this is not folded into proof-text-guards) for future-Craig-reading-the-archaeology

Out of scope

• App-side UI rendering for false_citations — separate card once content density justifies it (after we have ~10 entries, this surfaces as a "How to spot fabricated citations" Explore card; that is a future deliverable, not this one)
• Bulk-seeding additional false-citation entries — this card establishes the infrastructure and two seeds; subsequent enrichment is its own backlog
• Cross-linking from app screens (chapter views, panels) into false-citations — separate cards once the table is populated

Future cards this enables

Once this infrastructure ships, the following become small (XS) content-only cards:

• Pseudo-Jasher quotes (multiple in circulation)
• Misquoted Jeremiah 1:5 ("I knew you in your mother's womb" — actual: "Before I formed you in the womb I knew you")
• Pseudo-Gospel of Thomas sayings used in spiritual-but-not-religious content
• Misattributed C. S. Lewis / Bonhoeffer / Spurgeon quotations (out-of-canon but a known pattern)

Each subsequent entry uses the schema established here without further infrastructure work.