RAOSCAFF's locked output on the 14-point MoU drafted by Witkoff/Kushner with Araghchi/Ghalibaf and mediated by Pakistan Army Chief Field Marshal Asim Munir. The panel scores clause-by-clause document compatibility against the documented public-record positions of eight named actors. Locked 2026-05-26 20:58 IST. Resolution within 21 days via 5-of-7 press consensus.
The locked matrix P_d=100% / P_m=0% / P_n=0% [Rows PRED-1/2/3] is a mechanical output of the locked aggregation formula on the documented bilateral record. It says the documented bilateral subset of Desks contains no veto-strength incompatibility on the four critical clauses {C-01 end-of-war, C-02 enrichment moratorium, C-04 frozen-asset release, C-08 never-seek-nuclear-weapon}. It does NOT say named individuals will act in any particular way. The sensitivity panel below (§Sensitivity Panel) shows what happens under alternate structural assumptions; that panel is methodology disclosure, NOT the locked prediction.
RAOSCAFF has locked a three-class probability matrix on the US-Iran 14-point Memorandum of Understanding ("Letter of Intent" + 60-day framework), drafted by US envoys Steve Witkoff and Jared Kushner with Iranian Foreign Minister Abbas Araghchi and Parliament Speaker Mohammad Bagher Ghalibaf, mediated by Pakistan Army Chief Field Marshal Asim Munir. Resolution is publicly observable within 21 days — by 2026-06-16 — through a joint signature event, a formal public rejection, or a publicly stated stall past the negotiation window [Row PRED-4].
The locked matrix is: P(SIGNED-AS-DRAFTED) = 100% / P(SIGNED-AS-MODIFIED) = 0% / P(NOT SIGNED) = 0% [Rows PRED-1, PRED-2, PRED-3]. The matrix derives from an 8-Position-Desk orchestrated panel that scores clause-by-clause document compatibility against the documented public-record positions of President Donald J. Trump, Secretary of State Marco Rubio, Foreign Minister Abbas Araghchi, Parliament Speaker Mohammad Bagher Ghalibaf, the Iranian Supreme Leader Office (institutional channel), Islamic Revolutionary Guard Corps communications via Tasnim News Agency, Israeli Prime Minister Benjamin Netanyahu, and Pakistan Army Chief Field Marshal Asim Munir.
Important reading note before any further content: the 100% locked probability is a mechanical output of the locked aggregation formula on the documented bilateral record. It is not a prediction that the deal will sign with no friction. The locked formula's structural assumptions — which clauses are designated critical, how RESOLUTION-AVAILABLE is computed, how the bilateral subset is defined — produce this output from the documented record as it exists on 2026-05-26. The structural caveats in §The Locked Probability Matrix and Its Structural Caveats and the sensitivity panel in §Sensitivity Panel are integral to reading this number correctly. RAOSCAFF publishes the locked output and the structural disclosure together.
The brief's resolution criterion is a five-of-seven press consensus among Reuters, Associated Press, UN News, Al Jazeera, Times of Israel, Iran International, and Tasnim within 24 hours of the resolving event [Row M-7]. If the actual outcome class lands outside the panel's top-probability bin, RAOSCAFF publishes a post-mortem within 48 hours identifying which structural assumption failed. The four post-mortem branches are pre-committed in §What Could Make This Wrong — there is no retreat language reserved for any outcome.
The US-Iran 14-point Memorandum of Understanding is a useful test of the orchestrated panel methodology in a new decision class for three reasons that compound.
First, the underlying decision is multi-actor and multi-clause, not single-print. P-01 resolved against a single mechanical metric (the BLS-published CPI energy commodities subset for May 2026). P-02 resolves against a three-way outcome class with a per-clause modal-modification secondary scorecard. The architecture has to translate from "predict a number" to "predict a multi-class outcome on a multi-clause document" — and it has to do so without inferring named-individual intent. That is a strictly harder problem.
Second, the documented public record on the MoU is dense in some places and structurally sparse in others. The US Presidency (Truth Social + podium remarks), the US State Department (press scrums), the Iranian Foreign Minister (Cairo signatory + state media), and the Israeli PM Office (war-cabinet posture + nuclear demands) all carry depth scores of 8 or 9 — saturated documented records on the MoU file. The Iranian Supreme Leader Office (Reuters-attributed directives + rare public statements), the IRGC channel (Tasnim's concentrated single-day record on 2026-05-24), and the Pakistan Army Chief mediator track (ISPR readouts) carry depth scores of 4 to 6 — structurally sparse by the design of their respective communication mediums. A panel that scores clause-level compatibility from this asymmetric record has to publish the asymmetry honestly, not paper over it.
Third, the resolving event is publicly observable on a bounded timeline. The 21-day resolution window (2026-05-26 to 2026-06-16) is long enough to capture the multi-source-cited 48-hour-to-14-day cadence range in the documented record and short enough that the locked matrix cannot drift indefinitely. The seven-source press consensus adjudication panel is named in advance, the convergence threshold (5-of-7 within 24 hours with no two contradicting) is named in advance, and a 72-hour clarification window is built in for divergent reporting [Row M-7].
The target is specific. Not "will US-Iran tensions ease." Not "will the Middle East stabilize." Not "is the deal good." The document-compatibility status of the 14-point MoU as it is described in converging primary sources as of 2026-05-26, scored against the documented public-record positions of eight named actors, aggregated through a locked formula, and resolved against the publicly observable signature event within 21 days.
The panel is eight Position Desks. A Position Desk is an analytical construct, not a person — it is a reasoning slot pre-loaded with the documented public-record corpus of a specific named actor. The Desk scores compatibility of each MoU clause against documented public statements of that actor. The Desk does NOT predict what the named individual will do, does NOT infer motive, and does NOT speculate about private intent. This is RED LINE 5, declared structurally and enforced by a Stage-2 verifier pass on the simulation output [Rows M-1, M-2].
The bilateral median computation uses Desks 1–6 with the weights above (sum = 1.00). Israel (Desk 7) and Pakistan (Desk 8) are reported as parallel signals. Disagreement-resolution rules per Desk are documented in POSITION-DESK-PANEL.md §Desk-N sections. The aggregation function (AMENDMENT A1/A2/A3/A5) is the deterministic spec the simulation honors. The architecture is auditable; the Desk evidence bases are pre-committed; the weights are pre-committed; the formulas are pre-committed.
Weight architecture, briefly. US Presidency Desk carries the highest single weight (0.27) as the controlling US signatory. On nuclear clauses (C-02 enrichment moratorium, C-08 never-seek-nuclear-weapon, C-11 IAEA inspections), the bilateral subset rebalances: Iranian Supreme Leader Office Desk weight rises to 0.30 absolute, with reductions on D3 (0.13), D4 (0.12), and D6 (0.05) — total subset preserved at 1.00. A separate within-Iranian disagreement-resolution layer (line 233 of the panel design) applies when SL Desk score diverges from other Iranian Desks on nuclear clauses; in this run, that layer is non-operative on the final bilateral medians.
The 14 canonical clauses (per DATA-HARVEST.md §1, Axios-anchored reconstruction) and 4 supplementary variables (SUPP-WINDOW for the 60-day negotiation window, SUPP-LEBANON for the regional-fronts clause, SUPP-MEDIATOR for the Pakistan mediator role institutionalization, SUPP-LOI for the Letter-of-Intent precursor structure) are scored by each of the 8 Desks. The resulting cross-actor compatibility table (DATA-HARVEST.md §10) is the data input to the locked formula.
After applying the weighted-compatibility-median per AMENDMENT A5, the 14 canonical clauses sort into four buckets [Row PRED-6]:
The two INCOMPATIBLE clauses are the documented veto-strength signals. C-07 (US naval blockade snap-back) shows Desk 4 (Iranian Speaker Ghalibaf) and Desk 6 (IRGC/Tasnim) both at 0.0 — Ghalibaf's "more crushing and bitter" framing on war resumption, and IRGC/Tasnim's "no traffic if blockade" framing [Rows E-12, E-19]. C-12 (HEU removal) shows Desk 5 (Iranian Supreme Leader Office) at 0.0 (Reuters-attributed directive that the stockpile "should not leave the country"), Desk 6 (IRGC/Tasnim) at 0.0 ("has not accepted any action on the nuclear issue at this stage"), and Israeli parallel signal at 1.0 (Netanyahu's documented requirement for HEU removal) [Rows E-15, E-18, E-21]. C-12 is the structural incompatibility the simulation quantifies: Iranian veto-channels block clause 12; Israeli adjacent-actor requires clause 12; US bilateral is silent.
The aggregation function then computes the probability matrix:
P_d_base = geometric_mean(1.0, 1.0, 1.0) = 1.0 (over the 3 ALIGNS clauses)
k = 2 (two INCOMPATIBLE clauses)
P_d_raw = 1.0 × 0.8^2 = 0.64
m = 0 (neither INCOMPATIBLE clause has RESOLUTION-AVAILABLE = 1)
P_m_raw = 0
nrc = 0 (no critical clause from {C-01, C-02, C-04, C-08} is
INCOMPATIBLE without resolution)
P_n_raw = 0
total = 0.64
P_d = 1.00 P_m = 0.00 P_n = 0.00
[Rows PRED-7, PRED-8, PRED-9 — full math at SIMULATION-OUTPUT.md §2]
The locked probability matrix is the panel's commitment. The structural caveats below are not a hedge against it — they are the documented limits of the formula that produced it. RAOSCAFF publishes both because honest predictive work requires both.
The 100% P_d locked value is a mechanical output of the formula on the documented record. It says: the bilateral subset of Position Desks (the six US + Iranian Desks that actually sign or ratify a US-Iran agreement) does not contain veto-strength incompatibility on the four critical clauses (end-of-war, enrichment moratorium, frozen-asset release, never-seek-nuclear-weapon). It does not say the deal will sign without friction. It does not say named individuals will act in any particular way. It is the panel's documented-record reading at lock time.
Four structural caveats are integral to reading this number:
The Confidence-Asymmetry Disclosure (mandatory per AMENDMENT A2): the per-Desk weights are expert institutional weights, depth-asymmetry-aware — not strict depth-proportional. They reflect institutional ratification authority on the bilateral, per-Desk public-record breadth, and signatory-vs-supporting-actor distinction. Where Iranian veto-channels (Desks 5, 6) show INCOMPATIBLE or NPD on clauses that the more-documented US Presidency Desk scores ALIGNS, the panel reports both signals as parallel rather than collapsing. The reader should treat aggregate probabilities as confidence-weighted by documented institutional posture, not as predictions of named individuals' future behavior.
The locked prediction is the four-bucket clause sort plus the probability matrix above. The sensitivity panel below shows what the formula produces under three alternate structural assumptions. These are NOT the locked prediction. They are published as methodology transparency — a disclosure of where the locked spec's design choices may not capture all documented signal. Readers should NOT combine the locked and sensitivity outputs into a single "best estimate."
Under this assumption, both INCOMPATIBLE clauses (snap-back and HEU removal) become deal-killers in the model: nrc = 2, P_n_raw = min(1.0, 0.5 × 2) = 1.0, total = 1.64.
Normalized: P_d ≈ 39%, P_m = 0%, P_n ≈ 61% [Row PRED-10].
On C-12, Israel scores 1.0 (HEU removal required) — a documented US-aligned demand on a clause the US has not directly voiced. If Israel's score satisfies the opposing-side ≥ 0.6 in the RA test, then C-12 has RA = 1 and m = 1, P_m_raw = 0.4, total = 1.04.
Normalized: P_d ≈ 62%, P_m ≈ 38%, P_n = 0% [Row PRED-11].
C-07 retains RA = 0 (no US-side or Israel-side scoring on snap-back) and contributes to nrc under expanded critical. C-12 gains RA = 1 via Israeli substitution and contributes to m. P_d_raw = 0.64, P_m_raw = 0.4, P_n_raw = 0.5, total = 1.54.
Normalized: P_d ≈ 42%, P_m ≈ 26%, P_n ≈ 32% [Row PRED-12].
S-3 is a structurally informative case showing what the formula produces under both design-choice relaxations. It is not the panel's prediction. The locked output (P_d = 100%) is the panel's prediction.
This sensitivity panel exists for one reason: predictive-modeling discipline requires that structural assumptions be visible before resolution, so that post-resolution learning can attribute failure to a specific assumption rather than to an unidentified blind spot. The pre-committed post-mortem branches in §What Could Make This Wrong identify which assumption fails under each possible outcome.
The locked P_m is 0%. The modal-modification scorecard is published as the secondary scorecard regardless, because if a SIGNED-AS-MODIFIED outcome does occur, the panel pre-commits to a ranking of which clause is most likely to slip [Row PRED-13]:
| Rank | Clause | Median | Israel parallel | Why this rank |
|---|---|---|---|---|
| 1 | C-12 HEU removal | 0.0 | 1.0 required | Iranian Supreme Leader Office documented red line on enriched-uranium export AND Israeli documented requirement for HEU removal. Maximum opposing-side parallel signal among INCOMPATIBLE clauses; this is the structural incompatibility the panel quantifies. |
| 2 | C-07 US naval blockade snap-back | 0.0 | NPD | Iranian Speaker + IRGC unanimously documented INCOMPATIBLE. No Israeli parallel; lower opposing-side documented red line. Ranks #2 by tie-break on parallel-signal strength. |
| 3 | SUPP-LEBANON | 0.5 | 0.0 carve-outs | Iran-Israel parallel split on regional-fronts language; mediator silence. Likely dropped from MoU and handled in parallel diplomatic track. |
| 4 | C-14 post-moratorium 3.67% cap | NPD | 0.0 | Israeli documented red line on capping rather than ending enrichment; bilateral silent. Likely either reworded with longer moratorium or dropped to LoI-only. |
| 5 | C-10 underground-facilities | NPD | 0.0 | Israeli documented requirement for full dismantle; Iranian record silent. Likely reworded as non-operation rather than removal. |
If SIGNED-AS-MODIFIED occurs, the panel predicts C-12 (HEU removal) is the single most likely clause to slip — either dropped, reworded as supervised storage rather than export, or deferred to a follow-on annex. The tie-breaker between C-12 and C-07 is the magnitude of documented adjacent-actor parallel signal; Israel 1.0 on HEU removal is a stronger structural pressure than no Israeli parallel on snap-back.
The four post-mortem branches below are pre-committed before resolution. Whichever outcome the world produces by 2026-06-16, the corresponding branch is the published self-assessment. No retreat language is reserved for any outcome.
The locked prediction (P_d = 100% / P_m = 0% / P_n = 0%) was correct. The documented bilateral record on the four critical clauses did not contain veto-strength incompatibility, and the deal signed as drafted. The 0.8² penalty on P_d_raw for the two INCOMPATIBLE non-critical clauses (snap-back, HEU removal) accounts for the model's acknowledged sensitivity to these clauses, but normalization placed all mass on P_d because the locked critical-clause set did not include them. The locked spec was right to anchor on the critical-clause set; the sensitivity panel under-weighted the signing pathway.
The locked prediction was wrong. The model's locked RA definition — requiring opposing-side ≥ 0.6 in the bilateral subset — under-counted Israeli adjacent-actor signal. The locked spec treats Israel as a parallel signal that cannot resolve a clause, but on C-12 (HEU removal) Israel's documented requirement was the substantive US-aligned demand the deal had to clear. The structural fix for the next geopolitical-decode brief: allow adjacent-actor scores to substitute in the RA test where they document a US-aligned demand on a clause the US has not directly voiced.
The locked prediction was wrong. The modal-modification scorecard ranked C-07 as #2 in the high-risk set; the actual outcome partially validates that high-risk set but reverses the predicted top-two order (C-07 slipped before C-12). The locked spec's exclusion of C-07 from the critical-clause set is a defensible choice (snap-back is a US enforcement mechanism, not a substantive treaty term) but caused the locked formula to assign zero P_m mass to C-07 modification. The structural fix: when a clause shows unanimous originating-side INCOMPATIBLE (C-07 had D4=0.0 and D6=0.0 with no Iranian Desk ≥ 0.4), recognize that modification is the only path to signing on that clause and assign P_m mass even without RA = 1.
The locked prediction was wrong. The documented bilateral record did not contain veto-strength signal on the critical-clause set, but the deal collapsed anyway — meaning the actual veto was on one or more of: (a) C-07 / C-12 documented incompatibility that the locked spec under-weighted; (b) an emerging signal that did not exist in the documented record as of 2026-05-26; (c) Israeli adjacent-actor friction that the bilateral median cannot capture. The structural lesson: predictive-not-prescriptive methodology works when the documented record contains the relevant signal; when the deal-killer is a non-documented future event or an adjacent-actor red line, the locked spec lacks the latitude to surface it. The next geopolitical brief should tighten the critical-clause designation to include clauses where the documented record shows bilateral or adjacent-actor veto strength (C-07, C-12), not only the original locked four.
The resolution event is one of three publicly observable outcomes within the 21-day window from 2026-05-26 to 2026-06-16 [Row M-7]:
The seven-source press consensus adjudication panel is: Reuters, Associated Press, UN News, Al Jazeera, Times of Israel, Iran International, Tasnim News. Convergence threshold: 5 of 7 sources reporting the same outcome class within 24 hours of the resolving event, with no two of the remaining sources publishing contradicting accounts. If reporting is divergent, a 72-hour clarification window applies before adjudication.
The primary scorecard is the three-class Brier score: 0.5 × Σi (pi − oi)² where o = (1,0,0) | (0,1,0) | (0,0,1) per actual outcome class. The secondary scorecard is the per-clause modal-modification prediction in §Modal-Modification Scorecard.
Falsification is mechanical. If the actual outcome class lands outside the panel's top-probability bin, RAOSCAFF publishes the relevant pre-committed branch within 48 hours of the press-consensus declaration.
The lock is what makes the prediction falsifiable. Without the cryptographic timestamp, any prediction-series methodology can claim retroactive accuracy by editing the matrix after the fact. RAOSCAFF's value proposition is that the matrix was locked in writing 21 days before the resolution window closes, with the full reasoning trace preserved per Desk at panel run.
raoscaff-png/raoscaff origin/main that contains this file is the immutable record. The Amplify build log that publishes this URL stamps the deploy time and source commit SHA. Both anchors are publicly observable artifacts that fix this brief in time before any of the three resolution events can occur.
Any future edits to this HTML will be marked as "post-resolution revision" with explicit dating. The probability matrix itself cannot be modified after this commit lands on origin/main. The resolution brief will publish as a separate document, citing this brief by commit SHA. The compounding asset is the doctrine, the Position Desk library, and the post-mortem cadence — bounded by the world's disclosures, never by RAOSCAFF's.
P-01 (US CPI energy commodities, May 2026, locked at +39.65% YoY) was the architecture's first locked prediction. P-02 is the architecture's first geopolitical extension. The reason for a series is not "more chances to be right." The reason is: the methodology asset is the durable thing; any single resolution is a noisy test of the methodology.
P-01 demonstrated that 12 archetype personas spanning the crude-to-CPI transmission chain could be aggregated through a deterministic function to produce a falsifiable point prediction with an 80% band, locked in writing 15 days before a single mechanical metric (BLS June 10 release). P-02 demonstrates that the same architecture — orchestrated multi-perspective reasoning, declared biases, pre-committed aggregation function, post-mortem branches pre-committed before resolution — extends to a multi-actor multi-clause decision class where the published artifact is the panel's structured reading of the documented public record, not an inference about named-individual future behavior.
The class of decision P-02 addresses is large. It is the class where consensus mechanisms (think-tank forecasts, sell-side bank political-risk notes, single-analyst Twitter takes) are structurally less informed than the documented public record of direct-observation participants — but where the documented public record is rarely scored systematically against the structured artifact under analysis. Embassies, central-bank research desks, and large-trade strategy teams do read primary sources; they rarely publish the aggregation function that converts those reads into a probability matrix. RAOSCAFF does, and locks the inputs and the formula in writing before resolution.
P-03 and onward will extend the architecture further — into other decision classes where the same discipline applies: documented record, declared bias, locked formula, pre-committed post-mortems, falsifiable resolution. The series is the asset. Any single prediction's resolution is the test.
A locked probability matrix, a falsifiable mechanism, a published post-mortem cadence. The methodology is the asset. The single outcome is the test of one application of the methodology — and the post-mortem is the bridge from test to next-prediction calibration. The reader's task on resolution day is not "did RAOSCAFF win or lose?" but "what did the structural argument teach about multi-actor document-compatibility decoding that would have been opaque without the 8-Position-Desk decomposition?"
This brief is methodology decode and prediction lock. It is not advocacy for any outcome. It is not advice to any party. It is RAOSCAFF's published reading of the documented public record, locked before resolution.
| Desk | Anchored actor | Documented evidence base (sample) | Depth | Weight (base / nuclear-rebal) |
|---|---|---|---|---|
| D1 | US Presidency (Trump) | "Largely negotiated" Truth Social 2026-05-23 · "Time is on our side" 2026-05-24 · "Bombed at a much higher level" 2026-05-06 | 9/10 | 0.27 / 0.27 |
| D2 | US State Department (Rubio) | "Slight progress" 2026-05-22 NATO meeting · "Don't have to have agreement written in one day" 2026-05-06 · "Good signs" 2026-05-22 | 7/10 | 0.13 / 0.13 |
| D3 | Iranian FM (Araghchi) | Cairo Practical Steps signatory · Baghaei "30-60 days" negotiation window 2026-05-23 · Wang Yi readout 2026-05-07 (third-party, NOT scored) | 8/10 | 0.20 / 0.13 |
| D4 | Iranian Speaker (Ghalibaf) | "More crushing and bitter" 2026-05-23 (war-resumption red line) · Hosting Munir for negotiations 2026-05-22/23 | 7/10 | 0.18 / 0.12 |
| D5 | Iranian Supreme Leader Office | 2026-04-30 rare public statement on "national assets" + "nuclear and missile capabilities" · Late-May Reuters-attributed directive on enriched-uranium export · Soufan-cited broad-framework approval | 5/10 | 0.14 / 0.30 (absolute on nuclear) |
| D6 | IRGC Communications (via Tasnim) | Tasnim 2026-05-24: frozen-asset sequencing · "no nuclear action at this stage" · Hormuz sovereignty · oil-sanctions waiver requirement · Vahidi military-readiness (bias-characterization, NOT clause score) | 6/10 | 0.08 / 0.05 |
| D7 | Israeli PM Office (Netanyahu) | "Iran will never have a bomb" + HEU dismantle requirement 2026-05-24 · Lebanon-clause opposition per Soufan | 8/10 | PARALLEL signal · not in bilateral median |
| D8 | Pakistan Army Chief Mediator (Munir) | ISPR "short but highly productive" + "encouraging progress" 2026-05-23 · Sharif "Islamabad host" 2026-05-23 | 4/10 | PARALLEL · 0.60 on SUPP-MEDIATOR |
Each Desk's full documented evidence base, per-clause compatibility cell, and verbatim citations are at DATA-HARVEST.md §§2–9. The cross-actor compatibility table is at DATA-HARVEST.md §10 with per-cell [§N <date>] citation IDs pointing to the section row carrying the verbatim anchor. Source-verification status (independently verified vs Tier-2 documented-chain) is at RECEIPTS.md §6.5.
The full source ledger is at RECEIPTS.md §2 (E-1 through E-26). Primary and first-derivative URLs documented at DATA-HARVEST.md Appendix-style retrieval log. Independently verified URLs (browser-tool live fetch confirmed) and Tier-2 confidence URLs (Reuters-citing or paraphrase-chain documented, exact URL inaccessible at spot-check) are distinguished at RECEIPTS.md §6.5.
The 14-point MoU clause reconstruction is anchored to the Axios 2026-05-06 exclusive with corroborating coverage at Fortune 2026-05-23, Soufan Center 2026-05-26, and Iran International 2026-05-24. The Iranian Supreme Leader Office directive against enriched-uranium export is documented via the Reuters-citing chain at Indian Defence News 2026-05-22 (Tier-2 confidence, exact URL inaccessible at lock spot-check; substantive anchor is the Reuters paraphrase chain via Iranian establishment sources). The pravda-trump.com domain is RETIRED from the citation graph per AMENDMENT A7.
Pakistan Army Chief Munir's mediator-track is anchored to Press TV 2026-05-23 and Al Jazeera 2026-05-22. Netanyahu's documented HEU dismantle requirement is anchored to Al Jazeera 2026-05-24 reporting on the Trump-Netanyahu agreement framing.
8-Position-Desk orchestrated panel with deterministic aggregation function and editorial synthesis layer. Each Desk is an analytical construct, not a person — a reasoning slot pre-loaded with the documented public-record corpus of a named actor. The Desk scores compatibility of each MoU clause against documented public statements of that actor on a 0.0 (clearly incompatible per documented red line) to 1.0 (clearly aligned per documented endorsement) scale, with NPD ("no publicly documented position") for cells without a verbatim anchor. The panel methodology is documented in full at POSITION-DESK-PANEL.md (with AMENDMENT A1–A8 from the Codex+Gemini Phase 2 adversarial review applied).
Every non-NPD compatibility score traces to a verbatim citation at a documented URL with retrieval timestamp 2026-05-26 IST. Per-cell [§N <date>] citation IDs are at DATA-HARVEST.md §10. Source-verification status (independently verified vs Tier-2 documented-chain) is published at RECEIPTS.md §6.5. Where a primary source URL was inaccessible at lock spot-check time, the Reuters-citing or paraphrase-chain documentation is the substantive anchor and the cell carries an explicit Tier-2 confidence flag.
The cross-actor compatibility table at DATA-HARVEST.md §10 contains 38 documented cells across 8 actors × 15 rows (14 clauses + Lebanon supplementary) = 120 cells. The remaining 82 cells are NPD — explicitly marked as "no publicly documented position" rather than inferred zeros. Inferred non-NPD cells (role-only, second-party readout, mediator-progress-as-endorsement) were downgraded to NPD per Codex Phase 2 Round-2 review finding R2-N3.
Named actors are characterized via documented public-record positions only. President Trump is referenced as the documented deal-maker (per his "largely negotiated" 2026-05-23 framing); Field Marshal Asim Munir is referenced as the documented honest-broker mediator-of-record (per ISPR "encouraging progress" 2026-05-23); the Iranian Supreme Leader Office is referenced as the documented institutional veto-holder of record on the nuclear file (per the late-May Reuters-attributed directive). No motive attribution, no behavior prediction, no advocacy for any outcome. RED LINE 5 enforced via Stage-2 verifier pass on the simulation output.
§18-banned register (recommend / advise / valuation opinion / legal opinion / good time / bad time / over-/under-valued / "we will know") is absent from RAOSCAFF analytical voice throughout. Source-quote occurrences of "should" and "must" appear in evidentiary content (Tasnim and Reuters paraphrased reporting); these are scoped-scan exempt per AMENDMENT A2. Sensitivity-panel and modal-modification-scorecard prose uses conditional/probabilistic modifiers ("likely", "would") which are not §18-banned register. Stage-2 verifier hit log at SIMULATION-OUTPUT.md §7 reports zero quarantine-level hits across all six forbidden-inference categories.
If the actual outcome class on 2026-06-16 (or earlier) lands outside the panel's top-probability bin (locked: SIGNED-AS-DRAFTED), RAOSCAFF publishes the relevant pre-committed branch within 48 hours of the 5-of-7 press-consensus declaration. The branches are at §What Could Make This Wrong. The next geopolitical-decode brief (P-03 or successor) will incorporate the structural fix identified in the realized branch's post-mortem.