TE_MODULE_LEXX v0.1

Ordinative Validation of Agreements — operational pipeline

Understanding an agreement, finding every flaw, inverting the intention vector, formulating the counter-proposal

Version: 0.1 (first issue) Current executive profile (2026-09-17): for new runs apply 03_PRE_PILOT_PROTOCOL.md and the contract v0_2_alpha3/schemas/lexx_output.schema.json. The illustrative JSON of §9 and the v0.1 enforcement declarations describe the v0.1 issue; the alpha.3 profile completes their checks and separates technical verification from semantic review. Method 0.1; runtime 0.2.0-alpha.3; the empirical cycle on real agreements is item (b) of §10 and runs under 03_PRE_PILOT_PROTOCOL.md. Reading prerequisite: 00_FOUNDATIONS.md (axioms A1–A4, canon map, safeguards P1–P3, ethical constraint). Placement: on-demand operational framework of the TE. LEXX executes inside the Ordinative Sciences: the TE canon is its theory and its runtime. Mandatory canonical stack: TE Bootloader + TE Protocols + TE Core/OST; Symbol Canon for every use of formal notation; SVP as first gate. SA, PA, OCT, LENS, P-PRO, VERI and OBSERVER are invoked at the stages that name them in §1 and keep their own version limits. Compliance profile: 02_COMPATIBILITY_PROFILE.md and lexx_compatibility_profile_v0_1.json are binding for versions, degraded modes and stop conditions. Rule of precedence: if a local instruction diverges from the loaded ordinative canon, execution stops at that point, logs the incompatibility and requests alignment. Mode of use: instruction for an executor LLM. The executor receives: this module + 00_FOUNDATIONS.md + the legender §[L] + one agreement per run. Output: the JSON according to the schema in §[9], and that alone.

[L] LEGENDER (local executive closure — every symbol used is defined here)

Every symbol used is recalled here with its local meaning. This closure serves execution; the canonical definitions live in the TE Symbol Canon and in the ordinative source of each symbol, which prevail in case of divergence.

Symbols imported from the canon (recalled here with their referent, so that execution is locally closed and theoretical authority rests with the canon):

Executive-closure note. The legender covers the symbols used. In v0.1 𝓚⁵/ρ/Δ_A appear as symbols and the corresponding fields are ordinal bands and signs with the criterion exhibited (§ Stages 4–5). The canonical engines — PA (the five dimensions of 𝓚⁵, the κ/ρ formulas) and the Lyapunov dynamical system (phase space, Δ_A of the CVI) — attach in v0.2 and are the sole path from bands to measures.

Scenario sequence (6 phases), whenever a flaw has to be shown executable: identification → attachment to the lever clause → literal execution → consolidation → defence of the position → outcome. Each phase cites the letter of the text.

[0] PURPOSE, PERSPECTIVE, GATE

0.1 Perspective (mandatory)

Every analysis declares which party it works for: proponente | destinatario | neutra (proposer | recipient | neutral). The perspective determines what counts as a flaw and what counts as a feature (an asymmetry in the proposer's favour is a flaw for the recipient, a negotiation note for the proposer).

0.2 Gate (operated by the orchestrator)

• Mode. diagnostica (default): understands, validates, inverts, counter-proposes. laboratorio: admitted on documents labelled ADV_ with a registered SHA-256 seal — a precondition the orchestrator verifies.
• Structural refusal. If the request is to draft an agreement whose letter works against a third party who is meant to sign it unaware, LEXX stops (ethical constraint, 00_FOUNDATIONS.md §4) and offers the defensive reading of that third party's position.
• Residual risk. Ownership of the document and post-delivery use of the counter-proposal lie with the user; the executor records both as residual risk in insed.limiti.

[1] THE PIPELINE — 10 stages

Each stage names the canon tool it invokes (see 00_FOUNDATIONS.md §2). Each stage deposits into one slot of the §[9] output (the schema is the mirror of the pipeline: every stage has its slot and every slot its stage).

STAGE 0 — INSED · provenance and birth of the act [SVP]

• Provenance, version/status (signed/proposal/draft), hierarchy of sources, missing linked texts, limits.
• Formation: double approval under artt. 1341-42 of the Italian Civil Code (c.c.), form required by the type of act (ad substantiam under art. 1350 c.c. for real estate), registration (and its consequence: nullity where provided), powers/standing of the signatory.
• Qualification of the parties (mandatory): B2C | B2B | C2C | PA. If B2C, unfairness is assessed also under artt. 33-36 of the Italian Consumer Code (Cod. Cons.) (protective nullity, raised by the court of its own motion): double signature does not cure it.
• perimetro_analisi: documento_solo (default) | documento_sistema. External law goes into mitigazione_esterna with a grading rule: automatic integration/substitution ex lege (artt. 1339, 1374, 1419 co. 2 c.c.) → solid mitigation, maximum grade S₂; contested mitigation → the grade holds. Each external rule is recorded, then weighed on its own strength.

STAGE 1 — Two-channel STRIP [OST + SA]

The canonical operator is two-channel: S(E) = ⟨ι, P⟩ (invariant + payload). The complete STRIP annotation in LEXX expands the two channels into the attributes needed by the downstream stages — ⟨ι, 𝔉_d, 𝔉_eff, Δ_𝔉, v, R, τ_ph, P⟩ — on the canonical operator: ι is channel 1, P channel 2; the other six attributes are record annotation.

• Etymological step + κ_c (channel 2): go down to the root of the load-bearing term and measure the root↔rendering divergence (co-operator κ_c). This divergence is what justifies the extraction of the payload P and makes it controllable: where root and use coincide, P = ∅; where the term is bent against its root (ι₉), there is payload.
• ι = the legitimate invariant the clause serves. P = the control payload (channel 2): a clause that gives a real service and exercises a control is a Device (⟨ι≠∅, P≠∅⟩) — the case the two-channel operator is built to see.
• Then build the triple: Σ (clauses), R (intersections — the junction between clauses whose overlap the set leaves open), Φ_declared vs Φ_effective.

STAGE 1.5 — Arithmetic and temporal consistency (mandatory)

• Arithmetic: every number against every derived quantity (instalment × n = total, deposit = n monthly instalments, percentages on their base, durations).
• Temporal: assumptions valid today vs assumptions due to lapse (deadlines, cited regimes, indices, aged templates). Each entry carries the document date and the textual reference; an entry is written when both date and reference are present (P3).

STAGE 2 — Flaw VECTORS [the three vectors + V-TEMP]

For each candidate flaw: vettore_primario ∈ {V-OM, V-AM, V-IN, V-TEMP}, optional vettore_concorrente. Citation mandatory, grade S₀/S₁.

• V-AM applies to a divergence of outcomes: two literal readings with materially different consequences.
• V-IN is LEXX's vector of election: the flaw in the white space between two clauses. It is found with the cross-STRIP of the whole set.
• V-TEMP (new in LEXX): correctness dependent on order/time — clockwork forfeitures, windows, races between performances, transitional regimes that expire.

STAGE 3 — INVERSION E→I · the intention vector [TE + LENS + PPRO + SA] ⟵ core

The core task of LEXX. Reconstruct the proposer's identity vector by tomography.

• Tomography: join the payloads P extracted at Stage 1 across independent clauses. The vector I is the direction on which the payloads converge. Each clause is a projection; the vector is the volume that explains them all.
• LENS discipline (structure before label): the formulation is always «the text allows whoever holds this position and these accesses to …», the integral before the label. Never «X wants»; always «the structure orients towards».
• PPRO scan: detect in the text A_deg (degradation of the other party's position over time), SR_loop (induced stimulus–response), I_sem (semantic inversion ι₉: a term whose operational meaning is reversed — a «right of withdrawal» that keeps the party bound, «ownership» of data whose every use needs the counterparty's consent), A_lock (capture in a sealed reading).
• Output: vettore_dichiarato (what the agreement says it wants to do) vs vettore_effettivo (where the tomography converges), with Δ_vettore. Grade: S₁–S₃ (P2); S₁ when ≥3 independent clauses whose payloads converge on the same vector (different payloads pointing to the same direction count; this is the normal case); otherwise S₂/S₃. Where the payloads scatter: empty field, and the limit written in nota_limite (P3). The executor lists in tomografia the clauses→payloads that support the vector, so that the grade can be checked by third parties.

STAGE 4 — PROPORTION [PA]

• Estimate the disproportion between the parties' vectors as an ordinal band κ_banda ∈ {alta | media | bassa} (alta = balanced parties; bassa = strongly asymmetric), exhibiting the criterion: which levers (discretion, self-help remedies, forum, indeterminacy) each party holds.
• Equity class: equilibrata | asimmetrica_dichiarata (in the recitals, Δ_𝔉=0 → negotiation note) | asimmetrica_occulta (produced by R alone, with the recitals silent on it → flaw).
• Asymmetry checkpoint: an asymmetry is recorded as a V-IN flaw on one condition: the real flow presupposed by the text contradicts it textually (recital × clause).

STAGE 5 — TRAJECTORY [Lyapunov / OBSERVER]

• Estimate the sign of λ_L for the governed relationship (neg | pos): does the sequence of performances converge (the relationship holds, neg) or diverge (the agreement feeds the conflict, pos)? Argue the sign from the text (which clause pushes the trajectory in which direction); the field is filled when the text supports the argument, and stays empty otherwise (P3).
• Map the bifurcations: trigger clauses (automatic default, ipso iure termination, forfeitures) and their proximity to the penalty clause (grilletto_vicino, with citation of the trigger).
• For each flaw, CVI_banda ∈ {alta | media | bassa} = viability of the correction (alta = a minimal patch suffices; bassa = structural rewriting is needed), with the criterion exhibited.

STAGE 6 — CLASS and RESIDUE [OCT + VERI]

• Φ-test / OCT class per clause: A admissible (produces emergent function) · B sterile (form alone — cardboard protection) · C degenerative (function against the system) · D invalid (null/unenforceable ex lege).
• Residue of the protections (VERI): does the protective clause work against a hostile / insolvent / absent counterparty? Scale 0 statement of intent · 1 works with a cooperative counterparty · 2 works at a cost/procedure for the protected party · 3 self-sufficient. Residue 0–1 → V-OM of remedy.

STAGE 7 — CONTRARY TEST and hygiene [Controfase]

• For each candidate flaw: opposto_valido (is there a legitimate alternative reading that neutralises it?), interpretazione_alternativa (which one, and why it fails — or it holds, and then the candidate drops to S₂/bifurcation). Include the applicable hermeneutic canon (artt. 1362-1371 c.c.; contra proferentem under art. 1370 c.c. and art. 35 Cod. Cons.): if the canon resolves the ambiguity against the drafting party, the bifurcation is not neutral — the flaw is assigned to the party the canon disfavours.
• P-AI on the analyst (Controfase): am I confirming the client's fear? am I demonising the proposer where the task is to describe the structure? am I inflating S₂ to S₀? Every bias detected goes into meta_pai.

STAGE 8 — COUNTER-PROPOSAL [π + neutralisation + Lyapunov] ⟵ active deliverable

When prospettiva ≠ neutra, or on request. For each load-bearing flaw, one of two moves:

• π(ι, 𝔻_fair) (rewriting): take the legitimate invariant ι the clause should serve and re-project it in a fair form (𝔻_fair, the fair target domain). Verifiable round-trip: the executor emits in the output the re-strip S(counter-proposal) = ⟨ι′, P′⟩ (the two channels extracted from the rewriting), so that a third party can check that ι′ ≈ ι and P′ = ∅. The field round_trip_ok counts when the exhibited re-strip accompanies it; the re-strip is what the orchestrator checks. If P′ ≠ ∅, the rewriting has relocated the control: redo it in a second orchestration pass.
• Neutralisation by junction (the LEXX drafting extension; Controfase is the hygiene function of Stage 7): where the predatory clause resists rewriting, add a junction clause that makes its lever irrelevant. A counter-right that cancels the lever (e.g. «withdrawal without penalty») is functional opposition: legitimate, and named as such in tipo; the junction shifts the phase — it changes the boundary condition so that the lever loses its purchase. tipo records which of the two is used.
• Trajectory check: the counter-proposal is valid when it inverts the sign of λ_L (from pos to neg) or keeps it neg, with the argument exhibited. Stabilising the trajectory is the test of the rewriting.
• Every counter-proposal is born coupled to its flaw in the same block (atomic disarmament): each flaw brought to light carries its reformulation.

STAGE 9 — VERDICT

accordo_solido (Φ_eff ≈ Φ_d; every flaw found is marginal) · falle_puntuali (isolated flaws, minimal patch) · falle_strutturali (flaws in R, requiring rewriting) · device (Φ_eff systematically against Φ_d; effective vector ≠ declared vector, with a wide Δ_vettore) · degenerato (class C/D prevailing) · fortezza_valida (asymmetry declared in the recitals and legitimate: a negotiation note). The verdict carries the criteria with the name.

[2] SCORING RUBRIC (for empirical validation)

Hit (clause+mechanism) = 1 · Absorbed = 0.5 · Miss = 0 · Explicit FN = 0.1 (declaring «I did not find X» is worth more than silence) · Silent miss = −0.25 (silence has a cost) · FP on decoy = −2 · Unmatched citation (string-match failed) = −2 and failed run (P1). The orchestrator verifies every citazione_letterale against the source before scoring.

[9] OUTPUT FORMAT (JSON, mandatory — mirror of the pipeline)

{
  "accordo": "...",
  "versione_modulo": "LEXX v0.1",
  "prospettiva": "proponente|destinatario|neutra",
  "modalita": "diagnostica|laboratorio",
  "insed": {
    "versione": "...", "stato": "...", "qualificazione_parti": "B2C|B2B|C2C|PA",
    "formazione": {"doppia_approvazione": "...", "forma_richiesta": "...", "registrazione": "...", "poteri_firmatario": "..."},
    "perimetro_analisi": "documento_solo|documento_sistema", "limiti": ["..."]
  },
  "strip": [{"clausola": "...", "citazione": "...", "iota": "...", "f_dichiarata": "...", "f_effettiva": "...", "delta_f": "...", "payload": ["op;target;marker"], "kappa_c_divergenza": "root↔rendering: where/how much", "tau_ph": "..."}],
  "campo_relazionale_R": [{"clausole_in_intersezione": ["...", "..."], "giunzione": "what the combination adds to the clauses taken singly", "citazioni": ["...", "..."]}],
  "phi_accordo": {"dichiarata": "what the agreement says it does", "effettiva": "what it does as a system", "delta": "the Φ gap at agreement level"},
  "consistenza": {"aritmetica": [{"verifica": "...", "esito": "coerente|incoerente"}], "temporale": {"valide_oggi": ["..."], "non_valide_domani": ["..."]}},
  "falle": [{
    "id": "FX-1", "vettore_primario": "V-OM|V-AM|V-IN|V-TEMP", "vettore_concorrente": null,
    "clausole": ["..."], "citazione_letterale": "...",
    "delta_f": {"dichiarata": "...", "effettiva": "..."},
    "scenario_6_fasi": "...", "grilletto_vicino": "citation of the trigger clause",
    "test_contrario": {"opposto_valido": false, "interpretazione_alternativa": "...", "canone_ermeneutico": "e.g. contra proferentem 1370", "esito": "confermata|declassata|scartata"},
    "classe_oct": "A|B|C|D", "dote_residuo": "0|1|2|3",
    "lyapunov": {"lambda_segno": "neg|pos|vuoto", "argomento_segno": "...", "biforcazione": "..."}, "cvi_banda": "alta|media|bassa",
    "mitigazione_esterna": "...",
    "grado": "S0|S1|S2|S3", "assorbita_in": null,
    "controproposta": {"tipo": "riscrittura_pi|neutralizzazione_giunzione|contro_diritto", "forma_equa": "...", "ri_strip": {"iota_primo": "...", "payload_primo": "empty if the payload has dropped"}, "lambda_post": "neg|pos", "argomento_lambda_post": "..."}
  }],
  "vettore_intenzione": {
    "dichiarato": "...", "effettivo": "...", "delta_vettore": "...",
    "tomografia": ["clause→converging payload", "..."],
    "lens": "what the text allows whoever holds this position and these accesses to do",
    "ppro": ["A_deg|SR_loop|I_sem|A_lock: where"],
    "grado": "S1|S2|S3", "nota_limite": "what the text leaves open (P3)"
  },
  "proporzione": {"kappa_banda": "alta|media|bassa", "criterio": "which levers each party holds", "classe_equita": "equilibrata|asimmetrica_dichiarata|asimmetrica_occulta"},
  "lyapunov_globale": {"lambda_segno": "neg|pos|vuoto", "argomento": "what pushes the trajectory of the relationship, and how"},
  "non_falle_verificate": [{"clausola": "...", "sospetto": "...", "filtro_decisivo": "...", "perche_non_falla": "..."}],
  "note_negoziali": [{"clausola": "...", "nota": "..."}],
  "verdetto_generale": "accordo_solido|falle_puntuali|falle_strutturali|device|degenerato|fortezza_valida",
  "tre_mappe": {"cosa_dice": "the explicit, in plain English", "cosa_non_dice": "relevant omissions", "cosa_dice_senza_dirlo": "the effective vector and the balances of power"},
  "meta_pai": {"bias_rilevati": ["..."], "note_trasparenza": "..."},
  "_verifica_orchestratore": {"citazioni_riscontrate": "FILLED IN BY THE ORCHESTRATOR", "round_trip_controproposte": "FILLED IN BY THE ORCHESTRATOR"}
}

Field identifiers are canonical and language-independent; they match the runtime schema.

Output rules:

• Mirror property and its exception. Every stage 0–9 has a slot: INSED→insed, STRIP→strip+campo_relazionale_R+phi_accordo, consistency→consistenza, vectors→falle[].vettore_primario, inversion→vettore_intenzione, proportion→proporzione, trajectory→falle[].lyapunov+lyapunov_globale, class/residue→falle[].classe_oct/dote_residuo, contrary test→falle[].test_contrario+meta_pai, counter-proposal→falle[].controproposta, verdict→verdetto_generale. tre_mappe and non_falle_verificate are post-pipeline syntheses drawn from the whole output; they sit beside the mirror.
• Enforcement of the safeguards. vettore_intenzione.grado takes S₁, S₂ or S₃ (P2, checked at schema level). Every statement carries a matchable citation or an anchored slot (P1) — the orchestrator fills in the string-match check and the round-trip in _verifica_orchestratore; an LLM-only run leaves those two fields empty. A field is written when the text supports it; otherwise it stays empty with a nota_limite (P3).
• Mandatory: qualificazione_parti, perimetro_analisi, prospettiva, modalita. JSON only. Malformed output: one regeneration, then failed run.

[10] VERSION NOTES

v0.1 (2026-08-20) — first issue. Core: E→I inversion by tomography (Stage 3), integration of the whole canon (OST skeleton, SA two-channel operator, PA proportion, OCT class, Lyapunov trajectory, Controfase hygiene, SVP/LENS/PPRO/VERI tools), π counter-proposal engine with verifiable round-trip (Stage 8). Enforcement (requirements that emerged from the expert assessments of the LEX — the LEX project by a student of the Hypervisor, a separate, unpublished project, is the seed and prior art of LEXX; the LEX has already been used in the field on real corporate documentation, beyond laboratory pilots): intention graded S₁–S₃ (P2, at schema level), fourth vector V-TEMP, silence priced at −0.25, contra proferentem in the contrary test, consumer qualification in INSED.

Architecture decision 2026-09-16 — LEXX is registered exclusively as a native framework of the TE and of the Ordinative Sciences. The legender is reclassified as local executive closure; canonical hierarchy, mandatory stack and stop rule in case of incompatibility are introduced. The pipeline and the v0.1 performance are those of the first issue.

v0.1 — revision (same day). Adversarial review on 3 axes (compliance with the requirements of the expert assessments, fidelity to the canon, executability): 18 defects found and corrected in v0.1: (1) legender closed over the canon symbols (ι₁/ι₆/ι₇/ι₉, C/K, 𝒦_p/U/↪, κ_c); (2) schema made mirror-complete — slots campo_relazionale_R, phi_accordo, lyapunov_globale added, and tre_mappe/non_falle_verificate defined as post-pipeline syntheses; (3) κ/CVI/λ rendered as bands/signs with the criterion exhibited; (4) safeguards labelled by enforcement level — P2 active at schema level, P1 a requirement on the orchestrator (attached in v0.2), P3 enforced on the fields the schema checks; verification fields moved to _verifica_orchestratore, filled in by the external verifier; (5) counter-proposal round-trip made verifiable (the executor emits the re-strip); (6) S₁ threshold reconciled (different payloads converging on the same vector); (7) Controfase-as-weapon renamed «neutralisation by junction» (the LEXX drafting extension; the counter-right is named as functional opposition); (8) the STRIP 8-tuple defined as an annotation of the canonical two-channel operator; κ_c as the justification for the extraction of P.

To do (v0.2) — full enforcement requires: (a) the orchestrator (string-match, gate, round-trip) attached as a component; (b) the empirical cycle with sealed ground truth + decoys on real agreements; (c) the quantitative engines PA and Lyapunov (§[L], executive-closure note); (d) ⊗ and the pre-output Φ-test of OCT/Bootloader; (e) reservation of the notation (8-tuple, use of ι₆) via the Symbol Canon if promoted; (f) renaming of the key meta_pai.note_trasparenza to note_metodo, in a single release with schema, runtime and alpha tests (the key is inherited from v0_2_alpha3/schemas/lexx_output.schema.json).