# THE COUNCIL OF 900 — BENCH 2
## Families C · Aqidah & kalam (38 seats) · D · Fiqh & usul (56 seats) · E · Sirah & history (73 seats)
**167 MACHINE-SCHOLAR seats · 47 knowledge areas · sitting 2026-09-05**

> **EVERY SEAT ON THIS BENCH IS A MACHINE-SCHOLAR.** Not one is a human scholar, a mufti, a
> muhaddith or a mu'arrikh. This bench **documents**; it never adjudicates between schools and
> it issues no ruling. Where the tradition holds two positions, both are named and neither is
> crowned. Human review is invited and never waited for.
>
> **METHOD.** Every cell and every finding below was produced by opening the file and counting.
> Where a claim rests on no file it is marked **OPINION**. The v1 coverage matrix was a
> substring keyword probe; this bench re-ran it under word-boundary regexes and then read the
> artifacts, which reversed **31 of its 47 cells** in this bench's families.

---

## PART 0 — WHY v1 HAD TO BE RE-MEASURED

`build/P41-COUNCIL/coverage_matrix_v1.json` states its own law honestly: *"a probe miss is not
proof of absence"*. This bench found the **opposite** error to be the larger one — a probe
**hit** is not proof of presence. Three false-positive classes, each verified:

| Class | Example measured | Effect on v1 |
|---|---|---|
| **Nisba collision** — a Companion's tribal name read as a theological school | `build/P31-MIZAN/W7-M1/strata_layer.json` → all 4 `ash'ari` hits are `"attributor": "abu musa al-ashari رضي الله عنه"` — the Companion, not al-Ash'ari al-Mutakallim (d. 324 H) | *Ashari school* scored **PRESENT** |
| **Substring collision** — a technical term inside an unrelated word | `illa` (the usul term ʿillah) matched `istithna`, `Jibril`, `killa` → **13,947** ledger "hits" for *Qiyas & istidlal*; strict boundary regex returns **0** | *Qiyas & istidlal* scored TOUCHED |
| **Self-reference** — the charter that names the area counted as coverage of the area | `build/P41-COUNCIL/council_900_charter.json` is the top-ranked "built artifact" for *Siyar*, *Hijri calendrics*, *Rashidun*, *Archaeology*, *Contemporary fiqh academies* | 5+ areas scored on their own charter line |

A fourth, subtler class: **homonym across eras.** *History of science in Islam* scored 300
ledger hits on `bayt al-hikma` — every one of them from **P23 "Bayt al-Hikmah 2.0"**, the
waqf's own modern institute layer (`framework/_FRAMEWORK-40000.json` n=12001–12300), not
al-Ma'mun's ninth-century house.

**This bench's cell law (stricter than v1's, stated so it can be checked):**
- **PRESENT** — the canonical ledger gives the area a domain of its own **and** a built artifact carries area-specific rows.
- **TOUCHED** — the area appears only inside another discipline's layer (a source list, a tafsir cluster, a name-string), **or** the ledger names it but the built layer is a stub, a template or a declaration.
- **ABSENT** — no ledger domain and no non-incidental artifact.

---

## PART 1 — MATRIX CORRECTIONS (47 areas)

`✔` = v1 cell confirmed (basis may still be corrected) · `▲` = upgrade · `▼` = downgrade

### Family C — Aqidah & kalam (12 areas · 38 seats)

| # | Area | v1 | **Bench 2** | | Evidence (file · measure) |
|---|---|---|---|---|---|
| C1 | Tawhid studies | TOUCHED | **PRESENT** | ▲ | Ledger gives it two domains of its own — n 301–400 *Rububiyyah in the text*, n 401–500 *Uluhiyyah & worship*; built at `build/C1-ALLAH/D4-RUBUBIYYAH/`, `D5-ULUHIYYAH/` |
| C2 | Asma wa-sifat (doc) | PRESENT | **PRESENT** | ✔ | Ledger n 1–300 (Names/Attributes/Actions); `build/C1-ALLAH/D2-ATTRIBUTES/attributes_layer.json` = **21 rows** — present, but see F2 |
| C3 | History of kalam | TOUCHED | **TOUCHED (declared-only)** | ✔ | Sole basis is the AQ-01…AQ-12 table in `build/C1-ALLAH/D2-ATTRIBUTES/SOURCES.md`; **none of the 12 works exists on disk** (F1) |
| C4 | Ashari school (doc) | PRESENT | **TOUCHED (declared-only)** | ▼ | 0 ledger domains. All `ash'ari` matches in built layers = Abu Musa al-Ash'ari رضي الله عنه. AQ-04/05/08/09 registered, unacquired |
| C5 | Maturidi school (doc) | PRESENT | **TOUCHED (declared-only)** | ▼ | 0 ledger. Entire built footprint = 1 line each in six `SOURCES.md` files (AQ-06 *Kitab al-Tawhid*, AQ-10 *Nasafiyya*), both unacquired |
| C6 | Athari school (doc) | PRESENT | **TOUCHED (declared-only)** | ▼ | 0 ledger. Basis = AQ-01/03/07/11/12 in `C1-ALLAH/D3-ACTIONS/SOURCES.md`, all unacquired |
| C7 | Imami theology (doc) | TOUCHED | **TOUCHED** | ✔ | v1's 100 ledger "hits" were `ithna` inside `istithna` (P31 W4 S4) — a false positive. True basis: `00-CORPUS/hadith_shia.json` + al-Mufid's *Irshad*/*Amali* and Tusi's *Amali* held in `seerah_shia.json` |
| C8 | Mutazili heritage (doc) | ABSENT | **TOUCHED** | ▲ | **al-Zamakhshari — al-Kashshaf is held at 100.0% of 6,236 ayat** (`00-CORPUS/tafsir_manifest.txt`); its positions surface in the P17 consensus clusters |
| C9 | Eschatology (al-akhirah) | TOUCHED | **PRESENT** | ▲ | **A whole core pillar**: C6 THE HEREAFTER, 900 ledger points, nine domains, built at `build/C6-HEREAFTER/` (96 files). v1's largest single miss |
| C10 | Qadar & free-will (doc) | PRESENT | **PRESENT** | ✔ | C8 pillar, 900 points, `build/C8-QADR/` D1–D9 incl. D5-FOUR-LEVELS, D6-FREE-WILL, D8-MISCONCEPTIONS |
| C11 | Firaq & heresiography (doc) | TOUCHED | **TOUCHED (declared-only)** | ✔ | Basis = AQ-05 *Maqalat al-Islamiyyin*, registered and unacquired. No firaq layer exists |
| C12 | Comparative creed method | PRESENT | **TOUCHED** | ▼ | Ledger n 13001–13100 is named *'Aqidah & sirah as documented*, but the built layer (`build/P26-SCIENCES/D2-AQIDAH-SIRAH/aqidah_sirah_layer.json`) is **2 rows** anchored on 112:1 — no creed is compared with another |

### Family D — Fiqh & usul (18 areas · 56 seats)

| # | Area | v1 | **Bench 2** | | Evidence (file · measure) |
|---|---|---|---|---|---|
| D1 | Usul al-fiqh | PRESENT | **TOUCHED** | ▼ | Ledger n 12901–13000 is named *Fiqh & usul as documented*; the built layer is **2 entries on 1 verse** (`P26-SCIENCES/D1-FIQH/entries_normalized.json`). No usul work named anywhere |
| D2 | Hanafi school (doc) | PRESENT | **TOUCHED (incidental)** | ▼ | 0 ledger. No madhhab registry. Basis = incidental citations inside tafsir clusters |
| D3 | Maliki school (doc) | PRESENT | **TOUCHED (incidental)** | ▼ | 0 ledger. Basis = **al-Qurtubi's *al-Jami' li-Ahkam al-Qur'an* held at 100.0%** — a Maliki ahkam-tafsir on the shelf, never read as fiqh (F4) |
| D4 | Shafii school (doc) | PRESENT | **TOUCHED (incidental)** | ▼ | 0 ledger. Basis = quoted aphorism, e.g. `build/P17-TAFSIR/D1/consensus_103_2.json` → *"al-Shafi'i said: if people were to ponder this surah it would suffice them"* — an imam's saying, not a school position |
| D5 | Hanbali school (doc) | PRESENT | **TOUCHED (incidental)** | ▼ | 0 ledger. Basis = al-Sa'di's *Taysir al-Karim* held at 100.0%; no Hanbali fiqh text |
| D6 | Jafari school (doc) | PRESENT | **TOUCHED** | ▼ | v1's 300 ledger hits were `tusi` in the **P33 rijal** registry, not fiqh. True basis is stronger and stranger: **Tahdhib al-Ahkam (8.49 MB) + al-Istibsar (3.32 MB) — two of the Four Books — are sha-sealed on the shelf** (`00-CORPUS/acquired/shia_wave1/_META.json`), read only for rijal (F18) |
| D7 | Zahiri heritage (doc) | ABSENT | **ABSENT** | ✔ | 0 ledger, 9 files / 22 occurrences, 7 of them the charter |
| D8 | Qawaid fiqhiyyah | ABSENT | **ABSENT** | ✔ | 0 ledger; no maxims layer |
| D9 | Maqasid al-shariah | PRESENT | **TOUCHED** | ▼ | Ledger n 13101–13200 exists, but `P26-SCIENCES/D3-MAQASID/maqasid_layer.json` = **2 rows**, neither a maqsad; the classical layer contains **0** of `daruriyy`, `Shatibi`, `Juwayni`, `Ghazali`, `maslaha` (F19) |
| D10 | Ijma studies | TOUCHED | **TOUCHED** | ✔ | 0 ledger. The waqf's word *consensus* (P17) means tafsir agreement, not usul's ijma' — undeclared collision (F22) |
| D11 | Qiyas & istidlal theory | TOUCHED | **ABSENT** | ▼ | v1's 13,947 hits were the substring `illa`. Strict boundary regex over all 40,000 points: **0**. No analogy/istihsan/istishab layer |
| D12 | Fatwa literature history (doc) | PRESENT | **ABSENT** | ▼ | v1's 100 hits were `mufti` inside **P21 D1 "Retrieval & assistant — never a mufti"** — the waqf's refusal law, the exact inverse of the discipline (F20) |
| D13 | Ibadat mapping | TOUCHED | **TOUCHED** | ✔ | No ibadat registry. Nearest artifact is `build/C4-MUHAMMAD/D7-WORSHIP/` — 17 entries + 41 sirah-extract rows on **his ﷺ worship**, which is shamail, not an ibadat mapping |
| D14 | Muamalat mapping | TOUCHED | **ABSENT** | ▼ | 0 ledger, 19 files / 38 occurrences, no contract or riba layer |
| D15 | Family-law mapping (doc) | PRESENT | **ABSENT** | ▼ | v1's 113 hits were the English word `inherit` in the P12 successors registry. No nikah/talaq/faraid layer exists |
| D16 | Hudud documentation (doc) | TOUCHED | **ABSENT** | ▼ | 0 ledger; 36 files / 262 occurrences, all Qur'anic or hadith text, no register |
| D17 | Siyar & international law (doc) | TOUCHED | **TOUCHED** | ✔ | Basis corrected — real substrate exists at `build/P28-SIRAH-X/X11/narrative_locators.json`: the treaty formula `وادعهم` (6 works), *the Prophet's ﷺ written documents* `كتاب رسول الله` (13 works), delegations `وفد` (17 works) |
| D18 | Contemporary fiqh academies (doc) | PRESENT | **ABSENT** | ▼ | 2 files hold the phrase; both are `P41-COUNCIL/` self-reference |

### Family E — Sirah & history (17 areas · 73 seats)

| # | Area | v1 | **Bench 2** | | Evidence (file · measure) |
|---|---|---|---|---|---|
| E1 | Sirah source criticism | PRESENT | **TOUCHED** | ▼ | Ledger declares **X6 · The sources-criticism bench** (100 points) — `build/P28-SIRAH-X/` has **no X6 directory**. The nearest artifact, `X5/precedence_register.json`, carries `"rows": []` (F15) |
| E2 | Maghazi studies | PRESENT | **PRESENT** | ✔ | `build/P28-SIRAH-X/X2/maghazi_registry.json` — 15 battles probed, 15 located, per-work mention tables across 13 works. The bench's strongest built layer |
| E3 | Chronology of the revelation era | PRESENT | **TOUCHED** | ▼ | `X1/chronology_spine.json` — **21 periods, 21 of 21 `NEEDS-SOURCING`, zero dates** (F7) |
| E4 | Pre-Islamic Arabia | TOUCHED | **ABSENT** | ▼ | 0 ledger; no jahiliyyah layer; hits are incidental words in hadith text |
| E5 | Ansab (tribes & genealogy) | TOUCHED | **TOUCHED** | ✔ | **al-Baladhuri's *Ansab al-Ashraf*, 94,844 lines, is held** (`build/P28-SIRAH/corpus_inventory.json`) and read only for mention-counts; `P11-AHLBAYT/D3-TREE/tree_layer.json` = **1 row** (F17) |
| E6 | Historical geography of the Haramayn | PRESENT | **TOUCHED** | ▼ | No gazetteer exists. The nine `D9-ATLAS/atlas.json` files are **count summaries**, not maps — `C4-MUHAMMAD/D9-ATLAS/atlas.json` is 652 bytes of tallies (F23) |
| E7 | Rashidun era history | ABSENT | **ABSENT** | ✔ | Cell right, cause wrong: **Tabari's *Tarikh* (107,854 lines) and Ibn Kathir's *al-Bidaya* (153,379 lines) are held** — the Rashidun sources sit on the shelf with no ledger point to read them (F16) |
| E8 | Ahl al-bayt history | TOUCHED | **PRESENT** | ▲ | Ledger P11, 300 points, three named domains; `household_layer.json` names 5, both traditions' readings of 33:33 documented and unadjudicated — correct posture, thin body (F5) |
| E9 | Sahabah biography | TOUCHED | **PRESENT** | ▲ | Ledger P8 (600) + P10 (300) = 900 points across nine named domains. Cell is PRESENT; the body is **36 rows** (F5) |
| E10 | Isnad-based historiography method | PRESENT | **PRESENT** | ✔ | `build/P32-SHAHADAT/W-A/uncapped_links_v2_layer.json` — 33,868 rows over 4,987 verses, gate-asserted at `framework/p40_gates.py:G28` |
| E11 | Classical Islamic historiography | ABSENT | **TOUCHED** | ▲ | Tabari, Ibn Kathir, al-Ya'qubi, al-Mas'udi, al-Baladhuri, Ibn Abi al-Hadid all held and **used** — but only as sirah mention-counters (`X2`, `X10`, `X11`), never as historiography |
| E12 | Early-Islamic numismatics & epigraphy | PRESENT | **ABSENT** | ▼ | 0 ledger; no coin, inscription or papyrus register. Hits are the English word *inscription* in unrelated text |
| E13 | Archaeology of Arabia | TOUCHED | **ABSENT** | ▼ | 4 files / 12 occurrences; 9 of the 12 are the charter naming the seat |
| E14 | Hijri calendrics & dating | TOUCHED | **ABSENT** | ▼ | 6 files / 26 occurrences, 18 of them the charter. **No date-bearing field exists anywhere in the waqf** (F24) |
| E15 | Oral-tradition studies | PRESENT | **TOUCHED** | ▼ | 13 ledger hits, all the word `memoriz` in C3's hifz block (n 2401–2500). No talaqqi/mushafaha layer |
| E16 | Islamic civilization history | PRESENT | **TOUCHED** | ▼ | v1's 100 ledger hits are C2 n 1601–1700 *"Mercy in civilization"* — a mercy domain, not a history of civilization |
| E17 | History of science in Islam | TOUCHED | **TOUCHED** | ✔ | Basis corrected: all 300 `bayt al-hikma` hits are **P23 "Bayt al-Hikmah 2.0"**, the waqf's own institute layer, n 12001–12300 — an anachronistic homonym, not the Abbasid house |

**Bench 2 matrix totals — v1 → corrected**

| Cell | v1 (this bench's 47) | Bench 2 | Δ |
|---|---|---|---|
| PRESENT | 21 | **7** | −14 |
| TOUCHED | 20 | **24** | +4 |
| ABSENT | 6 | **16** | +10 |

31 of 47 cells changed or had their basis corrected. **v1 over-credited this bench's families
by 3× at the PRESENT level.** Not one correction implies dishonesty in the build — every
downgrade traces to the probe, and in every case the *artifact itself* was labelled honestly.

---

## PART 2 — THE FINDINGS (25, worst first)

**Centre-clause key** — (a) the Qur'an as the centre of research · (b) everything returns to
Allah ﷻ · (c) meaning derived through sirah / sunnah / hadith · **POLISH** = machinery, not centre.

| # | Finding | Area(s) | Cell | Sev | Centre | THE FILL |
|---|---|---|---|---|---|---|
| **F1** | **The aqidah/kalam corpus does not exist.** `build/C1-ALLAH/D2-ATTRIBUTES/SOURCES.md` §1 registers 12 works (AQ-01…AQ-12: *Wasitiyya*, *Tahawiyya*, *Ibana*, *Maqalat al-Islamiyyin*, al-Maturidi's *Kitab al-Tawhid*, *Nasafiyya*, *Maqsad al-Asna*, *Asas al-Taqdis*, *Dar' Ta'arud*, *Tadmuriyya*, *Lum'at al-I'tiqad*, *Sharh al-Tahawiyya*) at `00-CORPUS/aqida/aq-NN-*.json`. **That directory does not exist.** The seal file states it plainly: `_C1-D2-SOURCES-SEAL.sha256` line 2 = `# aqida/aq-01 through aq-12: pending acquisition`; the seal hashes 4 files, none of them aqidah. All 38 seats of family C rest on a shelf with no creed literature on it | C3–C6, C11 | TOUCHED (declared-only) | **HIGH** | **(b)** | **Acquisition.** 8 of the 12 are public-domain and on OpenITI/Shamela today — al-Tahawi's *'Aqidah*, al-Ash'ari's *Ibana* + *Maqalat*, al-Maturidi's *Kitab al-Tawhid*, al-Nasafi's *'Aqa'id*, Ibn Qudama's *Lum'a*, Ibn Taymiyya's *Wasitiyya* + *Tadmuriyya*. Same OpenITI pipeline that landed `shia_wave1` and `wave56` |
| **F2** | **The one domain that promised the kalam positions documents none of them.** `SOURCES.md` §6 clause 4 commits to *"the three major kalam positions — tafwid, ta'wil, ithbat — documented with their carriers and the evidence they adduced."* Measured on `attributes_layer.json`: occurrences of `tafwid` = **0**, `ta'wil` = **0**, `ithbat` = **0**, `Ashari` = **0**, `Maturidi` = **0**, `AQ-0` = **0**. The layer holds **21 rows** and **21 of 21** carry `"gloss_en": {"source": "UNSOURCED"}` | C2, C4–C6 | PRESENT | **HIGH** | **(b)** | **Nodes + acquisition.** After F1 lands: ~120 nodes on the C1 pillar — one row per attribute in the seed scope (al-'Ilm, al-Qudrah, al-Hayah, al-Iradah, al-Sam', al-Basar, al-Kalam, the face/hands/istiwa'), each carrying the three positions side by side with their AQ-file locus, exactly as §8 already specifies |
| **F3** | **Fiqh & usul — 56 chartered seats — resolve to 2 entries on 1 verse.** Ledger n 12901–13000 (100 points, panel-sealed). `P26-SCIENCES/D1-FIQH/entries_normalized.json` = **2 entries**; `graph_layer.json` = **2 nodes, 2 edges**, both pointing at `ayah:9:122`; `classical_layer_dataset.json` = 3 clusters of tafsir on 9:122. **Not one of the five madhahib is named. No usul work is named. `00-CORPUS/acquired/` holds zero fiqh and zero usul texts** (8 wave directories: gharib, irab_treebank, qiraat, shia_wave1, sunni_wave3, waqf_text, wave56, wave7) | D1–D9, D13–D18 | TOUCHED/ABSENT | **HIGH** | **(c)** | **Nodes + acquisition.** A P26 expansion of ~900 nodes (one 100-point domain per madhhab + usul + qawaid + maqasid + ibadat/muamalat mapping), fed by an acquisition wave: *al-Muwatta*, al-Shafi'i's *al-Risala* + *al-Umm*, *al-Mabsut*, *al-Mughni*, *al-Muhalla*, Ibn Rushd's *Bidayat al-Mujtahid*, al-Shatibi's *al-Muwafaqat* — all public domain, all on OpenITI |
| **F4** | **The ahkam-tafsir is already on the shelf and the fiqh domain never opens it.** `00-CORPUS/tafsir_manifest.txt` — **al-Qurtubi, *al-Jami' li-Ahkam al-Qur'an*, 100.0% of 6,236 ayat**, plus **Ibn Ashur *al-Tahrir wa'l-Tanwir* 100.0%** and **al-Razi *Mafatih al-Ghayb* 100.0%**. The Book's own legal and maqasidi commentary is held complete for every verse; `P26-SCIENCES/D1-FIQH` reads **one** verse of it | D1–D5, D9 | TOUCHED | **HIGH** | **(a)** | **An engine, not an acquisition.** An ahkam-extraction pass over the al-Qurtubi shard — the same 4-pass consensus engine already proven at `P17-TAFSIR/D1` — would give every ahkam verse a documented juristic layer with carriers named. The cheapest large win on this bench |
| **F5** | **The prosopographical spine is 57 persons.** Summed from every `*_layer.json` under the four people-pillars: P8 SAHABA 10+2+15+2+2+1 = **32**; P10 SAHABIYAT 1+2+1 = **4**; P11 AHL AL-BAYT 5+2+1 = **8**; P12 TABI'UN 10+2+1 = **13**. **57 rows against 1,500 declared, sealed ledger points.** Sharpest instance: ledger n 7301–7400 is titled *"The people of Badr — the registry of the 313+"*; `P8-SAHABA/D2-BADR/badr_registry_layer.json` holds **2 rows**, and its own text says *"the FULL 313+ name-by-name extraction is a dedicated pass, honestly AWAITED"* | E5, E8, E9 | PRESENT | **HIGH** | **(c)** | **Nodes + an engine.** Ibn Sa'd's *Tabaqat* (63,493 lines) and Baladhuri's *Ansab* (94,844) are held; the Badr roster is already located verbatim at char 685485 of the held Ibn Hisham. A roster-parser over that anchor yields 313 rows for roughly the cost of one wing. ~600 nodes across P8/P10/P11/P12 |
| **F6** | **The declared 59.4% "sirah witness" is co-occurrence, not placement.** `P28-SIRAH-X/X3/sirah_verse_sync.json` — 3,679 verses, and every record is exactly `{"quote_clusters": N, "in_delimiter_ground_truth": bool}`. **No verse carries a sirah event, a year, a place or a locus.** The layer's own law concedes it: *"per-passage joins (which PAGE of which work) are the next engine on this spine."* The owner's law is *meaning derived through the sirah*; today the machine can say a verse **is quoted in** a sirah work, never **where in the life it sits** | E1, E3, E6 | TOUCHED | **HIGH** | **(c)** | **An engine.** The passage-join the layer already names: verse → work → page-marker → the narrative unit around it. The held corpora carry 1,606 page-markers in Ibn Hisham and 3,862 in Ibn Sa'd — the addressing scheme exists. ~300 nodes |
| **F7** | **The chronology spine carries no chronology.** `X1/chronology_spine.json` — 21 periods (*Before the mission* … *the Farewell*), and **21 of 21 read `NEEDS-SOURCING`**. Zero years, zero dates, zero sourced content. 100 ledger points (n 26401–26500) stand `selftest-green` over it | E3, E14 | TOUCHED | **HIGH** | **(c)** | **Nodes + acquisition-free reading.** The years are in the held Ibn Sa'd, Tabari and Ibn Kathir. ~200 nodes: one sourced entry per period with its work + page-marker, then the maghazi registry (X2, already built) supplies the Madinan years |
| **F8** | **The standing gate batteries do not touch a single pillar of this bench.** Read directly: `framework/p31_w9_gates.py:26` = `glob.glob(os.path.join(B31, "*", "*.json"))` where `B31 = build/P31-MIZAN` → 91 files. `p31_w9_gates2.py:18,43,71` — same root plus one P30 file. `p40_gates.py` asserts over P33, P34, P35, P36, P37, P32 only. **G18–G31 execute zero assertions over C1–C8, P8, P10, P11, P12, P26, P28 or P28-SIRAH-X.** The preliminary audit cites three GREEN batteries as the machine evidence for the whole waqf | all 47 | — | **HIGH** | POLISH | **An engine.** A G32–G36 battery scoped to the core pillars: row-count parity between each `*_layer.json` and its `graph_layer.json`, an `UNSOURCED`-ratio ceiling per domain, a `NEEDS-SOURCING`-vs-green-mark cross-check, each with a negative control that must fire (the p40 pattern, already law) |
| **F9** | **A gate that audited nothing was recorded GREEN.** `reviews/PANEL-P26-PILLAR.md` line 13: *"`prepanel_gate.py ../build/P26-SCIENCES` → **PRE-PANEL GATE GREEN: 0 json + 0 pages audited across G1-G9** ✅ VERIFIED"*. Zero files audited, nine gates green. The same review then verifies `entries_normalized = 2` against `graph_layer nodes = 2` and records **"✅ MATCH"** — internal consistency confirmed, cardinality never asked | D1, D9, C12 | TOUCHED | **HIGH** | POLISH | **An engine.** Make an empty audit set a RED, not a GREEN — a gate whose denominator is zero must fail. One assertion in `prepanel_gate.py`. Then a standing **points-per-row** ratio in every panel's evidence table |
| **F10** | **10,000 nodes stand on one evidence string.** `framework/_TRACKER-JOURNAL.jsonl` — a single evidence line (*"PANEL-SEALED: PX1/PX2/PX3 sittings … reviewed P33-P37 + the frame"*) closes **10,000 points** spanning eight pillars: P35 (2,500), P33 (2,000), P34 (1,500), P37 (1,500), P36 (1,000), REPAIR-2 (800), RESERVE (500), P28-COMPLETION (200). From the isnad-historiography seat this is a single chain asked to carry 10,000 mutun — the tradition's own name for that load is *gharib*, and it is graded accordingly | E10 | PRESENT | **MED** | POLISH | **An engine.** Per-domain evidence stamping: the closing evidence line records the artifact **that domain** produced, not the sitting that reviewed the batch. The journal is append-only, so this is a forward law + a re-stamp wave |
| **F11** | **500 points defined as landing-earned were closed by a review.** Ledger n 39501–40000, `"domain": "RESERVE-EARNED"`, `"purpose": "500 earned only by landings sealed during this run"`. Final status = `panel-sealed`, and **all 500 carry the F10 evidence string** — a panel sitting. No landing is named in the evidence of any of the 500 | E10 | — | **MED** | POLISH | **Re-stamp, not nodes.** Either name the landing that earned each block (waves 1–6 did land, sha-sealed, and would carry ~450 honestly) or re-open the balance as reserve. Cost: one journal pass |
| **F12** | **X6, X12 and X13 are declared and unbuilt.** The ledger declares *X6 · The sources-criticism bench*, *X12 · Sirah pages & education*, *X13 · Gates & panel* at 100 points each. `build/P28-SIRAH-X/` contains X1, X2, X3, X4, X5, X7, X8, X9, X10, X11, X14, X15 — **no X6, X12 or X13 directory**. 300 declared points, no artifact directory | E1 | TOUCHED | **MED** | **(c)** | **Nodes.** X6 is the family-E seat's own bench — Ibn Ishaq's isnad-less reports, al-Waqidi's split reception, the mursal maghazi. ~300 nodes; the material is entirely in the held works |
| **F13** | **The sources-criticism register is an empty table.** `X5/precedence_register.json` — `"rows": []`, and 4 `seeded_positions` each marked `NEEDS-VERBATIM-RESOURCING`, including the discipline's central pair (*al-Waqidi: matruk in hadith, relied upon in maghazi*; *Ibn Ishaq without isnad: the scholars' split stances*). The schema is exact and the register is empty | E1 | TOUCHED | **MED** | **(c)** | **Nodes.** Four rows, verbatim-sourced. The carriers are named in works already held — Ibn Hajar's and al-Dhahabi's verdicts on al-Waqidi are quoted inside al-Suhayli and Ibn Sayyid al-Nas, both on the shelf |
| **F14** | **A persons table built on unresolved name-strings.** `P8-SAHABA/D3-NARRATORS/narrators_layer.json` — 15 rows under a `"name"` field. Measured, they are **≤9 persons**: Abu Hurairah/AbuHurayrah · Ibn 'Abbas/Abdullah ibn Abbas · A'ishah/Aisha, Ummul Mu'minin · Ibn 'Umar/Abdullah bin 'Umar · Anas bin Malik/Anas · Jabir/Jabir bin 'Abdullah — **six pairs** — plus a bare `"Abdullah"` (245 lines) that resolves to no one. `reviews/PANEL-P8-PILLAR.md:196` disclosed **four** pairs and passed it; the disclosure undercounts by two and misses the unresolvable string entirely | E5, E9, E10 | PRESENT | **MED** | **(c)** | **An engine.** The id-join law the waqf already states (`build/P37-GRAPH/education.json`: *"Edges join on verse-keys and report-ids, never on name-strings or translations — a graph built on strings inherits every homonym as a false edge"*) and already gates for verses (`framework/p40_gates.py` G31, with a firing negative control), applied to persons: rename the field `attribution_string`, add a `person_id` resolved against the P33 registries, keep the strings visible |
| **F15** | **P33's 2,000 points rest on 15 KB.** `build/P33-RIJAL/` — 10 files, 15 KB total, against 2,000 declared ledger points. `rijal_registries.json` gives **Tusi's *Rijal*: `text_blocks_parsed` = 69** for a 584,931-byte work of several thousand narrator entries. The file is exemplary in its honesty — it renamed the field away from "entries", declares the proxy's error direction per work, and states *"NO entry-count claim stands until the per-work roll parsers run"*. The defect is not the layer; it is that 2,000 points are sealed above it | E5, E10, D6 | PRESENT | **MED** | POLISH | **An engine.** The per-work roll parsers the file itself names — Tusi's `(ع)`-coded line entries, Kashshi's own numbering (already parsed: 1,147 blocks / 433 sections). Until then, the 2,000 points should read `partial`, not sealed |
| **F16** | **Two of Islam's greatest histories are held and no seat reads them as history.** `P28-SIRAH/corpus_inventory.json` — **al-Tabari, *Tarikh al-Rusul wa'l-Muluk*: 107,854 lines** and **Ibn Kathir, *al-Bidaya wa'l-Nihaya*: 153,379 lines**, plus al-Ya'qubi, al-Mas'udi, al-Baladhuri. Both are read **only** for sirah mention-counts (`X2`, `X10`, `X11`). The ledger has **0 points** for Rashidun-era history and **0** for classical historiography | E7, E11 | ABSENT / TOUCHED | **MED** | **(c)** | **Nodes only — the acquisition is already done and sealed.** ~400 nodes: a Rashidun spine (the four caliphates, the ridda, the futuh, the first fitna) documented from Tabari with both traditions' readings preserved side by side, exactly as P11 already handles 33:33 |
| **F17** | ***Ansab al-Ashraf* — 94,844 lines — against a family tree of 1 row.** `P11-AHLBAYT/D3-TREE/tree_layer.json` = **1 row**, whose text is the line itself (the Prophet ﷺ — Fatimah and Ali — al-Hasan and al-Husayn). Baladhuri's genealogical compendium and Ibn Sa'd's tabaqat are both held. No nasab, tribe or clan registry exists anywhere in 40,000 points | E5 | TOUCHED | **MED** | **(c)** | **Nodes + an engine.** A nasab parser over Baladhuri's own genealogical formulae (`بن … بن …` chains) gives Quraysh's clan structure from the source. ~200 nodes on P11/P8 |
| **F18** | **The acquisition shelf is asymmetric across the madhahib.** `00-CORPUS/acquired/shia_wave1/_META.json` — *Tahdhib al-Ahkam* (8,488,344 B) and *al-Istibsar* (3,317,887 B), **two of the Four Books**, sha-sealed under CC BY-NC-SA. Against them: **zero** primary text of the Hanafi, Maliki, Shafi'i, Hanbali or Zahiri schools. The imbalance is a by-product (they were acquired for rijal), but a waqf that documents fiqh from this shelf would document one school's substrate and no other's | D2–D6 | TOUCHED | **MED** | **(c)** | **Acquisition.** The F3 wave restores symmetry. Until it lands, `P26-SCIENCES/D1-FIQH/SOURCES.md` should carry an explicit shelf-asymmetry declaration — the waqf's own honest-gap law applied to the source list |
| **F19** | **The maqasid domain contains no maqsad.** `P26-SCIENCES/D3-MAQASID/maqasid_layer.json` = **2 rows**: *"the maqasid thread"* (pointing at C2's mercy architecture) and *"the sciences explorer"*. Its `classical_layer_dataset.json` (25 editions, 9 clusters on 21:107) contains **0** occurrences of `daruriyy`, `Shatibi`, `Juwayni`, `Ghazali` and `maslaha`. The five daruriyyat — din, nafs, 'aql, nasl, mal — are enumerated nowhere in 40,000 points | D9 | TOUCHED | **MED** | **(c)** | **Nodes.** ~100 nodes on P26/D3: the five daruriyyat, each with its transmitted carriers (al-Juwayni → al-Ghazali → al-'Izz → al-Shatibi → Ibn Ashur), positions documented and none crowned. Ibn Ashur is **already held at 100.0%** |
| **F20** | **v1 mapped a discipline onto its own refusal.** *Fatwa literature history* scored PRESENT on 100 ledger points that are **P21 D1 "Retrieval & assistant — never a mufti"** (n 11401–11500) — the waqf's law against issuing fatwa. A probe that reads a prohibition as coverage will read every refusal law as a discipline | D12 | ABSENT | **MED** | POLISH | **A matrix law.** Refusal-law text (`never-a-mufti`, `never adjudicated`, `the waqf adopts none`) must be excluded from coverage probes by construction, and `P41-COUNCIL/*` excluded from its own denominator. One filter in `p41_council.py` |
| **F21** | **1,339 canonical ledger titles carry a baked-in truncation ellipsis, and 67 carry a grammar defect.** Measured over `_FRAMEWORK-40000.json`: titles containing U+2026 = **1,339** (C4 117, C1 91, C6 91, C3 78 …); titles containing `" the the "` = **67**, beginning at **point n=3: *"Hash & seal the the Names of Allah ﷻ sources"***. A template variable was truncated and the ellipsis was written into the append-only ledger. The waqf's adab law gives every Name its due honor; a malformed title on the Names-of-Allah point is the one place this cannot stand | C1, C2 | PRESENT | **MED** | **(b)** | **A repair wave.** The ledger is append-only by law, so this is a versioned v2.0.1 correction of the title field with the diff recorded — not a silent edit. 1,339 titles, mechanical |
| **F22** | **"Consensus" carries two meanings and the waqf declares neither.** `build/P17-TAFSIR/D1/consensus_*.json` uses *consensus* for **tafsir agreement across editions above a threshold of 3**. In usul, *ijma'* is a source of law. `P26-SCIENCES/D1-FIQH/fiqh_documented_layer.json` then writes *"9:122's consensus"* inside a **fiqh** layer. A reader of the fiqh pages meets the word in its juristic sense; the machine means the other one | D10, D1 | TOUCHED | **MED** | POLISH | **A glossary law.** Rename the fiqh-layer usage to `attested-agreement`, and add a glossary entry — already a built surface (`education_pack.json` carries `glossary`) — declaring that the waqf's *consensus* is never ijma' |
| **F23** | **Nine files named "atlas" and not one is a map.** `C4-MUHAMMAD/D9-ATLAS/atlas.json` = 652 bytes of tallies (`{"shamail": {"manner": 647, "speech": 49 …}}`). There is no place registry, no coordinate, no Makkan or Madinan topography anywhere in the build. `X10`/`X11` locate 15 narrative anchors by Arabic string only (`شعب أبي طالب`, `دار الأرقم`, `الطائف`) | E6 | TOUCHED | **LOW** | **(c)** | **Nodes.** A ~100-node Haramayn gazetteer keyed to those anchors — each place with its transmitted name, its works+page-markers, and its verses. The strings are already located; only the registry is missing |
| **F24** | **The waqf holds no date.** No layer in 40,000 points carries a hijri or CE date field: `X1` has 21 periods and no year, `X2` has 15 battles and no year, and the *Hijri calendrics & dating* probe returns 6 files / 26 occurrences of which 18 are the charter. The waqf cannot express *"this verse, that year of the mission"* — the ordering primitive of both sirah and asbab al-nuzul | E14, E3 | ABSENT | **LOW** | **(c)** | **An engine.** A dating register: hijri↔CE conversion, the tradition's own dating conventions (and its disputes — the year of Badr is agreed, the year of the Isra' is not), with every date carrying its transmitting work. Small, and it unlocks F6 and F7 |
| **F25** | **The two-axis declaration is inherited, and the second half of the ledger is structurally thinner.** Corroborating the preliminary audit's finding (`points_with_explicit_axis_fields: 0 / 40000`) with an independent measure: **26,400 points carry the keys `kind`+`title`+`pillar_name`; the remaining 13,600 (n 26401–40000) carry none of them** — only `pillar`, `domain`, `block`, `purpose`. The charter's *"every node declares both axes"* is true by structural inheritance in the first 66% and by inheritance-through-a-thinner-record in the last 34% | all 47 | — | **LOW** | **(b)** | **A stamping pass.** Two fields per point — `axis_reverence` and `axis_research` — written mechanically from pillar placement, plus back-filling `kind` and `title` on the 13,600. Mechanical, no judgment, and it converts an inherited claim into a measured one |

---

## PART 3 — WHAT THIS BENCH DID **NOT** FIND

Stated because a bench that reports only defects is itself an I-1 violation.

| Observed | Measure |
|---|---|
| **No fabricated source.** Every work named in an inventory that this bench opened, exists at the declared size | 17/17 sirah works in `corpus_inventory.json`; 21/21 tafsir editions in `tafsir_manifest.txt`; 5/5 + 4/4 acquisition files in `shia_wave1`/`wave56` `_META.json` with sha256 |
| **No hidden gap.** In every thin layer found above, the layer's own text declares the thinness first — *"honestly AWAITED"*, *"NEEDS-SOURCING"*, *"NO entry-count claim stands"*, *"SLOT DECLARED - never faked full"* | F5, F7, F13, F15 |
| **Never-a-mufti holds.** This bench found no ruling issued, no school crowned, no verdict on a disputed matter anywhere in family C or D | `P26-SCIENCES/D1-FIQH/classical_layer_dataset.json` `"difference_preserved": []` with agreement labelled *what commentators recorded*; `P11/D1` documents **both** traditions' readings of 33:33 unadjudicated |
| **Sectarian symmetry is real where it was built.** Shii sources are held, sealed, licensed and used, and their absence from a shard is declared rather than hidden | `consensus_2_255.json` `_meta.honesty_notes`: *"All 27 editions in shard for 2:255 are Sunni; no Shia editions"* |
| **The maghazi registry is genuine scholarship.** 15 battles located by the works' **own naming**, both bare and `ghazwat`-prefixed, with a stated law that a zero is a naming-convention fact rather than an absence | `X2/maghazi_registry.json` — Badr in 13 works, Uhud in 11, Khandaq in 10 |
| **The prior sirah bench's structural blocker was closed.** `reviews/COUNCIL-BENCH4-SIRAH.md` found neither *al-Shama'il* nor *Dala'il al-Nubuwwa* held; both are now on the shelf, sha-sealed | `00-CORPUS/acquired/wave56/` — `Shamail.txt` 275,315 B, `Dalail.txt` 5,296,275 B |

---

## PART 4 — THE FILL, CONSOLIDATED

| Fill class | What it is | Findings served |
|---|---|---|
| **Acquisition — 1 wave** | 8 public-domain aqidah works (AQ-01…AQ-12 minus the 4 rights-unclear) + 7 public-domain fiqh/usul works, via the proven OpenITI pipeline | F1, F2, F3, F18 |
| **Nodes — ≈2,900** | fiqh/usul ~900 · people-registries ~600 · Rashidun & historiography ~400 · X6/X12/X13 ~300 · sirah passage-joins ~300 · nasab ~200 · chronology ~200 · attributes-positions ~120 · maqasid ~100 · Haramayn gazetteer ~100 | F2, F3, F5, F7, F12, F13, F16, F17, F19, F23 |
| **Engines — 6** | ahkam-extraction over al-Qurtubi · sirah passage-join · person-id resolution · rijal roll-parsers · dating register · G32–G36 core-pillar gate battery | F4, F6, F14, F15, F24, F8 |
| **Repairs — 4** | ledger title repair (1,339 + 67) · per-domain evidence re-stamp · RESERVE-EARNED re-stamp · axis-field stamping | F21, F10, F11, F25 |
| **HUMAN seats — 5** | **(1) a mutakallim** to receive the tafwid/ta'wil/ithbat layer before it publishes; **(2) a faqih/usuli** for the madhhab mapping — the never-a-mufti law makes this a documentation seat, not a fatwa seat; **(3) a muhaddith-rijali** for person-identity, which the waqf's own law reserves to scholars (*"the identification is scholars' work"* — `rijal_registries.json`); **(4) a sirah/maghazi historian** for the chronology's disputed years; **(5) a Shii-tradition scholar** so that the Imami material — already held — is documented by its own tradition rather than about it | F2, F3, F5, F7, F14 |

**The bench's own reading of the centre test.** Of 25 findings, **12** touch clause (c), **4**
touch (b), **1** touches (a), and **8** are POLISH. Family C's gap is clause (b) — the
tradition's own articulation of how everything returns to Allah ﷻ is registered and unheld.
Family D's gap is clause (c) — the sunnah-derived meaning of the Book's rulings, with the
substrate (al-Qurtubi) already on the shelf. Family E's gap is the **hinge of clause (c)**:
the waqf can prove a verse is quoted in the sirah, and cannot yet say **where in the life** it
stands. Every one of these is a gap in the meaning-derivation layer, not in the evidence
layer — which corroborates the preliminary audit's central measure (evidence 59–80% of the
Book, meaning 1.4–2.3%) from three independent disciplinary vantages.

**Nothing found by this bench contradicts the waqf's honesty law.** Every thin layer named
itself thin before this bench opened it. What the bench found is that the **ledger's marks sit
above those declarations**: 1,500 points sealed over 57 persons, 100 sealed over 2 fiqh
entries, 100 green over 21 unsourced periods, 2,000 sealed over 15 KB. The artifacts tell the
truth; the point-counts do not yet carry it.

---

BENCH 2 VERDICT: SEAL WITH GAPS — the build is honest everywhere this bench measured and no source, count or claim was found fabricated, but three of the waqf's forty-seven chartered disciplines here (aqidah/kalam, fiqh/usul, Islamic historiography) stand on a shelf that holds no text of their own, and 31 of this bench's 47 coverage cells were over-credited by the v1 probe — so the seal must carry the gap register in Part 2, and the first human seats named in Part 4 must be sought before any of these areas is described as covered.

*— THE COUNCIL OF 900, BENCH 2 · 167 MACHINE-SCHOLAR seats · families C, D, E · 2026-09-05.
Documented, never adjudicated. Never a mufti. Human review invited, never waited for.*