# Verixa Wave 1 — D1 Patent-vs-Trade-Secret Election + Disclosure-Coverage Check (Repo-Verified)

> **CONFIDENTIAL — Verixa Patent Attorney Workbench output. Attorney-Client Privileged. Not a final legal opinion. Not filed. Internal use only.**
> Date: 2026-05-29 · Scope: **provisional filing only**, Wave 1 (Band A) · Perimeter: **Verixa only** (Vimal, Founder)
> Repo audited: `verixaai/verixa` @ `dev-vimal-audit-2` / `3bd2fe47` (read-only).

## Model Selection Note
- Model used: Claude Opus (this session) · Effort: high (repo audit = xhigh-adjacent; symbol-level only, no source disclosed)
- Why: patent-vs-trade-secret election + spec↔repo verification are §3 high-effort, implementation-dependent
- Output status: **preliminary — attorney-review-ready.** Not a filing decision, not a legal opinion.
- Stronger model review recommended: Yes for the D1 election sign-off and counsel confirmation of each match
- Reason: trade-secret election and enablement are founder + counsel calls

## Scope discipline (provisional-only)
A provisional secures a priority date for **what it discloses** (§9 / §10(4)). Priority attaches only to disclosed matter; matter held back cannot later claim this priority date. Therefore the only substantive pre-filing content decision is **D1 (what to disclose vs hold)**, and the only quality gate on the disclosure is **coverage** (does the provisional disclose the elements the complete-spec will claim, and is that disclosure repo-true?). Claims, §3(k), FTO, double-patenting → complete-spec (deferred).

---

## PART 1 — Repo audit (spec ↔ code truth)

Symbol-level verification on `dev-vimal-audit-2`. "Confirmed" = the disclosed mechanism is implemented in code.

| Patent | Disclosed core mechanism | Repo evidence | Verdict |
|---|---|---|---|
| **P01** | Ordered pipeline; TDAL `SET LOCAL`; closed-by-default route registry; hash-chained audit; validated-model gateway; HITL e-sig; agentic panel | Component symbols present (TDAL, llm_audit_log ×15, gateway, e-sig, panel modules) | **Confirmed** |
| **P02** | `llm_audit_log` 3-hash chain; `model_registry` qualification state machine; `agent_qualification_*` | `qualification_status` ×15, `model_registry` ×12, `agent_qualification_*` ×4, llm_audit_log ×15 | **Confirmed** (spec honestly states H2/H3 are *administrative services*, matching `preCallCheck`/`assertQualified` API posture) |
| **P02b** | Per-inference `output_type` (advisory/deterministic) on `ai_requests`; prohibited-field write-prevention; advisory evidence pack | `'advisory'` string literals only (audit-resource-registry, DQG); **`output_type` / `outputType` field NOT found** | **⚠ Partial — disclosed data structure unverified.** Mechanism concept present; specific field as disclosed unconfirmed → prophetic or verify |
| **P03** | Per-tenant LLM hash chain: systemPromptHash + userPromptHash + responseHash; prompt-version pinning; GENESIS sentinel; chain verification | system/user/response hash ×6/6/4; `GENESIS` ×9; llm_audit_log ×15 | **Confirmed** |
| **P04** | Tenant isolation; SHA-256 e-sig signature chain; advisory-lock serialisation; TOTP MFA; 5/15 lockout; SoD; live authority revalidation; E7 panel | SoD ×60, TOTP ×10, lockout ×19, advisory-lock ×10 | **Confirmed** |
| **P05** | Context-gate profiles; home-site sentinel; scope enforcement; `SET LOCAL`; migrations 092/132/259 | migrations 092/132/259 present; `home_site` in context-gate/service.ts | **Confirmed** |
| **P09** | Cross-tenant hash-chain replay resistance; advisory locks; GUC isolation; E4a evidence (URS-36) | `packages/backend/test/integration/cross-tenant-leak.spec.ts` **exists**; URS-36 refs present; advisory locks ×10 | **Confirmed** (E4a condition from 05-29 memo is satisfiable — evidence committed) |
| **P10** | Agent registry trigger_event_types; qualification gate; `DOMAIN_RELEVANCE_MATRIX`; deliberation phases; 3 conflict types; composite scoring; 8 escalation triggers | `DOMAIN_RELEVANCE_MATRIX` in `composite-scoring.service.ts:15`; trigger_event_types ×5; conflict types ×4; panel_conflicts/escalations ×9; deliberation_responses ×7 | **Confirmed — values match verbatim** (see D1-1) |

**Repo audit conclusions:**
1. Seven of eight Band A provisionals are **repo-true** at the disclosed-mechanism level.
2. **P02b** is the one mismatch: the disclosed `ai_requests.output_type` field is not found in code (only `'advisory'` string literals). Either (a) implement/confirm the field before relying on it, or (b) keep it as a prophetic embodiment and ensure at least one *implemented* element (advisory labeling, prohibited-field gate) carries the independent disclosure. Provisionals may contain prophetic matter, but repo-grounded matter is stronger for priority + later support.

---

## PART 2 — D1 Patent-vs-Trade-Secret Election (Founder-Memo §7 hold-back list × Band A)

For each §7 hold-back item: is it **disclosed** in a Wave-1 provisional, and what is the election? Rule: disclose the **mechanism** (needed for enablement + it's the moat); hold the **production tuning constants** unless disclosure is required to enable the claimed mechanism.

| # | §7 hold-back item | Disclosed in Wave 1? | Repo finding | Election (attorney recommendation — founder/counsel to confirm) |
|---|---|---|---|---|
| **D1-1** | **Agentic-panel domain-weighting tuning** (DOMAIN_RELEVANCE_MATRIX values) | **YES — P04 + P10 disclose the actual values** | **Values match production code verbatim** (`composite-scoring.service.ts:16`) | **For the June 15 provisional: holding is LAWFUL — see Part 2A.** §10(4)(b) best-method binds the *complete* specification, not the provisional; §11(2) gives the eventual mechanism-claim the provisional's priority date so long as the claim is "fairly based" on the disclosed *mechanism* (which it is, without the exact values). So you may file the provisional disclosing the weighting **mechanism + illustrative values** and **hold the production matrix** for now. **The irreversible election is deferred to complete-spec (FD+12mo):** if you then claim the weighting invention, §10(4)(b) likely compels disclosing the production matrix as the "best method… known to the applicant" (withholding it → §64(1)(h) revocation exposure, a live ground in *Lava v Ericsson* 2024). **Recommended provisional action:** keep the mechanism, replace the verbatim production table with clearly-labelled *illustrative* values, and **correct the inaccurate "representative" label.** Decide patent-vs-secret on the matrix at complete-spec, not now. |
| **D1-2** | **Confidence-threshold tuning curves** | Partially — P10 discloses fixed thresholds (severity-disagreement = 4 pts; ≥3 triggers→QP; relevance ≥0.8; 24h SLA; "confidence below threshold") | Thresholds are coded constants | Same §10(4)(b)/§11(2) timing logic as D1-1. **Provisional:** disclose the multi-trigger escalation *structure* (the moat); genericise exact numbers ("a configurable threshold") — lawful to hold at provisional stage. **Complete-spec:** if these constants are the best method for the claimed escalation invention, they must be disclosed then. "Confidence below threshold" is already generic — fine. |
| **D1-3** | **Drift-detection heuristics** | Mechanism only — P01 §6.13 / P02 §7.3 disclose version-change classification → recommendation (no-change/minor/major → proceed/requalify/block) | Rule structure in code; no secret curve disclosed | **Safe to disclose as-is.** What's disclosed is the *rule framework*, not a proprietary tuning curve. If a more sophisticated production drift heuristic exists, it is NOT disclosed and should stay secret. **No change needed.** |
| **D1-4** | **Agentic-panel internals (orchestrator)** | Mechanism — P01 §6.8, P04 E7, P10 disclose phases, conflict types, escalation tiers | Implemented (agentic-panel module) | **Disclose mechanism (it's the moat).** Hold back: any non-disclosed orchestration optimisations. Current disclosure is mechanism-level — acceptable. |
| **D1-5** | **Handover-bridge generator** | Mechanism — P01 §6.15 (intent classifier, payload transformer, cascade-rule registry, max depth) | Implemented (handover-bridge module) | **Disclose mechanism.** Hold back: specific cascade-rule sets / classifier tuning if competitively valuable. Current disclosure is structural — acceptable. |
| **D1-6** | **Mira prompt templates + citation corpus** | **Not disclosed** (P01/P02 reference prompt-version *pinning* by hash/version id, not template text) | Prompt text not in specs | **Correctly held.** Pinning the *hash/version* (not the text) is exactly right — protects priority on the audit mechanism while keeping the prompt text secret. **No change; preserve this discipline.** |
| **D1-7** | **Training data sources** | **Not disclosed** | n/a | **Correctly held.** No action. |

**D1 bottom line (law-grounded — see Part 2A):** For the **provisional**, the only required edits are cosmetic-but-important: in P04/P10 replace the verbatim production matrix/thresholds with clearly-labelled *illustrative* values and delete the inaccurate "representative" wording. Holding the production tuning at provisional stage is **lawful** (§10(4)(b) binds the complete spec, not the provisional) and does **not** forfeit priority for the mechanism claim (§11(2) "fairly based"). The **irreversible patent-vs-trade-secret election is a complete-spec decision (FD+12mo)**, not a June-15 blocker. D1-3/4/5 disclose mechanism only (safe). D1-6/7 are correctly held back.

---

## PART 2A — Legal grounding (Indian Patents Act 1970 + corpus case law)

*All citations are to documents in `indian_patent_laws_and_rules_bundle/` and `indian_patent_case_laws/`. No authority outside the corpus is relied on. Counsel must confirm current text on ipindia.gov.in before filing.*

**Statute (Patents Act 1970, consolidated — `…/downloads/1970_The_Patents_Act_Act_39_of_1970_consolidated_IndiaCode.pdf`):**

- **§10(4)(b)** — "Every **complete specification** shall… disclose the **best method of performing the invention which is known to the applicant** and for which he is entitled to claim protection." *(verbatim)* → The best-method duty attaches to the **complete spec**, not the provisional. This is why holding the production matrix at provisional stage is lawful, and why the irreversible election lands at FD+12mo.
- **§10(4)(a)** — complete spec must "fully and particularly describe the invention and its operation or use and the method by which it is to be performed." **§10(5)** — claims must be "fairly based on the matter disclosed."
- **§11(2)** — a complete-spec claim takes the **provisional's filing date as its priority date *only if the claim is "fairly based on the matter disclosed"* in the provisional**; **§11(3A)** — 12-month priority on the matter first disclosed. → This is the legal test for the Part 3 coverage check: priority is secured only for matter the provisional discloses. Disclose the mechanism broadly now.
- **§64(1)(h)** — a granted patent may be **revoked** if the complete spec "does not sufficiently and fairly describe the invention and the method by which it is to be performed… **or that it does not disclose the best method of performing it which was known to the applicant**." *(verbatim)* → The teeth behind D1-1/D1-2: if the production matrix is the best method of a *claimed* weighting invention and is withheld at complete-spec, the patent is revocation-exposed.
- **§64(1)(i)** — revocation where a claim "is not fairly based on the matter disclosed"; **§64(1)(k)** — subject not patentable (the §3(k) hook); **§64(1)(l)** — prior **secret use in India** before the claim's priority date. → If the production tuning is commercialised secretly before a later claim's priority date, that can itself imperil the later claim.
- **§3(k)** — "a mathematical or business method or a computer programme *per se* or algorithms" are not inventions. (Eligibility substance is delegated to the 3(k) Expert; relevant here only to how the provisional frames technical effect.)

**Case law (corpus):**

- **Lava International v Ericsson, Delhi HC 2024** (`india/official/2024_Lava_…WIPOLex.pdf`) — directly engages **§3(k) for computer-related inventions and the "technical effect" test** (¶¶ on "algorithms… with a technical effect," the computer-programme-*per se* limb, and the dispute over whether the technical-effect exception reaches algorithms), **and** a **§64 insufficiency / best-method / fair-description revocation challenge** to the suit patents (the "do not sufficiently or fairly describe the invention or the best [method]" ground). → Confirms in a current Delhi HC ruling that (i) CRIs survive §3(k) only on a demonstrated **technical effect/contribution**, and (ii) **insufficiency and best-method are live, litigated revocation grounds**. Both discipline Verixa's complete-spec drafting and the D1 election. *(Pleadings/propositions cited; counsel to read the court's holdings in full before relying on outcome.)*
- **Biswanath Prasad Radhey Shyam v Hindustan Metal Industries, SC 1978** (`india/official/1978_…WIPOLex.pdf`) — an improvement on what is known must be "**something more than a mere workshop improvement**" and must "independently satisfy the test of… an inventive step"; obviousness to the skilled person negatives inventive step. → Inventive-step discipline for the **combination/composition claims (P01/P06)** and the **deliberation claims (P10)**: the GxP-event-type weighting + multi-trigger escalation must be framed as a non-obvious technical combination, not an obvious application of known multi-agent voting.
- **Novartis AG v Union of India, SC 2013** (`india/public_mirrors/2013_Novartis_…`) — Indian patentability exclusions are construed **strictly against the patentee** (in re §3(d)). → By analogy, §3(k) framing must be conservative; do not rely on AI/compliance usefulness.
- **Monsanto Technology v Nuziveedu Seeds, SC 2019** (`india/official/2019_…WIPOLex.html`) — subject-matter exclusion (§3(j)) and **revocation operate at the claim level**. → Reinforces the §64(1)(k) discipline: each claim stands or falls on its own patentability.

**Net legal opinion (provisional-only, June 15):** Filing the Band A provisionals as disclosure-rich documents is sound. Under §11(2)/§9 the provisional secures priority for every mechanism it "fairly… discloses," so **disclose each invention's mechanism broadly now** (the Part 3 coverage check confirms all eight do). Under §10(4)(b)/§64(1)(h) the **best-method duty does not bite until the complete spec**, so the production tuning constants (D1-1/D1-2) may lawfully be **held back from the provisional** — replace with labelled illustrative values. The genuine, irreversible patent-vs-trade-secret election on those constants is a **complete-spec decision at FD+12mo**, to be taken with counsel: claim the weighting invention (and disclose best-method tuning) **or** keep the tuning secret (and claim around it). Nothing in the corpus authorities makes the production values mandatory at provisional stage.

---

## PART 3 — Disclosure-Coverage Check (§11(2) "fairly based" test: does the provisional disclose what the complete-spec will claim?)

| Patent | Core elements the complete-spec will need | Disclosed in provisional? | Coverage verdict |
|---|---|---|---|
| **P01** | Pipeline + TDAL + route registry + hash audit + validated-model gateway + HITL + panel; integrated-combination embodiment (§6.11) | Yes — §§6.1–6.15 + §6.11 integrated embodiment + method + worked embodiments | **Strong.** Extended embodiments (drift, panel-degradation, handover) carve-out language present to protect later separate filings. |
| **P02** | H1 inline audit chain; H2/H3 as administrative services (honest framing) | Yes — §§7.2–7.5; honest "administrative service / optional integration" language | **Strong.** Framing avoids overclaiming inline gating. |
| **P02b** | Output classification; prohibited-field gate; advisory evidence pack; per-tenant RLS | Yes at mechanism level — §§6.2–6.10 + claim seeds | **Adequate but verify D1/output_type.** Ensure ≥1 independent, *implemented* element anchors the disclosure (advisory labeling + prohibited-field gate are implemented). |
| **P03** | 3 independent hashes; prompt-version pinning; GENESIS; append-only trigger; per-tenant RLS; chain verification | Yes — detailed abstract + body; claims deferred | **Strong — cleanest.** Discloses the Ascent-distinguishing elements, so complete-spec can claim around C2. |
| **P04** | E3 isolation + E6 e-sig ceremony + E7 deliberation + handover integration | Yes — full abstract + body | **Strong, but E7 overlaps P10** (demarcation = complete-spec issue; not a provisional blocker). |
| **P05** | Context-gate profile store; home-site sentinel resolution; scope enforcement; pre-flight validator; RLS | Yes — body + formal Claims 1–7 | **Strong + repo-true** (migrations confirmed). |
| **P09** | Cross-tenant boundary; replay resistance; advisory-lock serialisation; GUC isolation; LLM sub-chain | Yes — body + formal Claims 1–7; E4a evidence exists | **Strong.** Overlaps P03 (LLM sub-chain) — demarcation at complete-spec. |
| **P10** | Domain-weighting matrix; routing; qualification gate; 4 phases; 3 conflict types; composite scoring; 8 triggers; 3-tier escalation | Yes — body + formal Claims 1–7 | **Strong + repo-true.** Subject to D1-1/D1-2 election; overlaps P04 E7 (demarcation at complete-spec). |

**Coverage conclusion:** All eight Band A provisionals disclose enough to support their intended complete-spec scope. Only **P02b** has a coverage caveat (verify the disclosed data structure is implemented or rely on the implemented elements).

---

## Red Flags (provisional-only scope)
1. 🟠 **D1-1 / D1-2 (production tuning disclosed verbatim in P04/P10).** Pre-filing action is limited and cosmetic: substitute labelled *illustrative* values for the verbatim production matrix/thresholds and remove the "representative" mislabel. Holding the production tuning is **lawful at provisional stage** (§10(4)(b) binds the complete spec). The **irreversible patent-vs-trade-secret election is a complete-spec decision (FD+12mo)** under §10(4)(b)/§64(1)(h) — not a June-15 blocker.
2. 🟠 **P02b spec↔repo mismatch** (`output_type` field not found). Prophetic matter is permissible in a provisional and does not defeat priority, but at complete-spec the claim must be **fairly based (§11(2))** and **enabled / best-method (§64(1)(h))** — so verify the field is implemented or anchor P02b on its implemented elements (advisory labelling + prohibited-field gate).
3. 🟠 **Filing gates G1 (applicant identity) + G2 (inventorship/Form 5)** remain open — unchanged; still the true blockers.
4. 🟢 **P09 E4a** — resolved: `cross-tenant-leak.spec.ts` exists in repo.
5. 🟢 **Repo prior-art bar (OI-9)** — closed: private, never public.
Deferred (NOT provisional blockers): claims posture, C2/Ascent, §3(k), FTO, P04↔P10 / P03↔P09 demarcation.

## Recommended next actions
1. **Provisional edit (Drafter via Router — Sub-Workflow E):** in P04/P10 replace the verbatim production matrix/thresholds with clearly-labelled *illustrative* values and delete the "representative" wording. Keep the weighting/escalation *mechanism* fully disclosed (priority under §11(2) attaches to the mechanism). No founder election needed for the provisional.
2. **Calendar the real D1 election for complete-spec (FD+12mo):** founder + counsel decide patent-vs-trade-secret on the production tuning under §10(4)(b)/§64(1)(h) — claim the weighting invention (disclose best-method tuning) or claim around it (keep tuning secret). Add to the docket now.
3. **Verify P02b `output_type`** field is implemented (DIY grep / engineer) or re-anchor P02b disclosure on its implemented elements (§11(2) fair-basis + §64(1)(h) enablement at complete-spec).
4. **Clear G1 (file as Vimal individual) and G2 (inventorship confirmation, §64(1)(b)/(m) exposure).**
5. **Frame technical effect now** in each provisional's disclosure to support the later §3(k) technical-effect test (*Lava* 2024) and inventive-step (*Biswanath Prasad* 1978) — mechanism + technical effect, not compliance/business result.
6. File Band A as India provisionals on/before 2026-06-15.

## Source Ledger
| Source type | Source / path | Date | Claim supported | Limitations |
|---|---|---|---|---|
| Drafted spec | P01/P02/P02b/P03/P04/P05/P09/P10 Form 2 | 2026-05-27→29 | Disclosure structure, §7-item disclosure status, claim posture | Bodies read at section + targeted-window level |
| Repo (read-only) | `dev-vimal-audit-2` @ `3bd2fe47` | 2026-05-29 | Spec↔code truth; matrix value match; symbol presence | Symbol-level audit; not a line-by-line code review |
| Repo file | `composite-scoring.service.ts:15` | 2026-05-29 | DOMAIN_RELEVANCE_MATRIX production values = disclosed values | — |
| Repo file | `packages/backend/test/integration/cross-tenant-leak.spec.ts` | 2026-05-29 | P09 E4a evidence exists | Test presence; behavior not re-run |
| Strategy | Founder Memo §7 trade-secret list (via Portfolio Index) | 2026-05-28 | D1 hold-back items | Memo not re-read in full this pass |
| Statute | Patents Act 1970 §10(4)(a)(b), §10(5), §11(2)/(3A), §3(k), §64(1)(h)(i)(k)(l) — `…/1970_The_Patents_Act_…IndiaCode.pdf` | consolidated; verify current | Best-method binds complete spec; "fairly based" priority; revocation grounds | Counsel to confirm current text on ipindia.gov.in |
| Rules | Patents Rules 2003 consolidated to 15 Mar 2024 — `…/2024_Current_Eversion_Patents_Rules_2003_…_fulltext.md` | to 15-Mar-2024 | Form 3/5 periods; procedure | 2025 Amendment Rules (GSR 865E) also in corpus — verify |
| Case law | *Lava International v Ericsson*, Delhi HC 2024 — `india/official/2024_…WIPOLex.pdf` | 2024 | §3(k) technical-effect for CRIs; §64 insufficiency/best-method as live revocation grounds | Propositions/pleadings cited; counsel to read full holdings |
| Case law | *Biswanath Prasad v Hindustan Metal*, SC 1978 — `india/official/1978_…WIPOLex.pdf` | 1978 | Inventive step; "more than a workshop improvement" | Applied to P01/P06/P10 combination framing |
| Case law | *Novartis v UoI*, SC 2013; *Monsanto v Nuziveedu*, SC 2019 — corpus | 2013/2019 | Strict construction of §3 exclusions; claim-level revocation | Cited at INDEX-supported level |
| Doctrine | `Aeonn_Founder_Doctrine/CLAUDE.md` | 2026-05-29 | Verixa perimeter, fail-safe phrases, Tier routing | Incorporation Unknown |

## Final Caveat
> This is a Verixa patent attorney workbench output, not a final legal opinion. Qualified patent counsel and local counsel where required must verify law, facts, repo evidence, claim scope, filing authority, prosecution strategy, FTO conclusions, trade-secret risk, publication clearance, and jurisdiction-specific requirements before action.

---
*Verixa Patent Attorney — D1 Election + Disclosure-Coverage (repo-verified) · 2026-05-29 · Perimeter: Verixa only · CONFIDENTIAL — Attorney-Client Privileged · Not Filed.*