蒋道理｜商标异议·无效申请推理引擎（SJ-IRAC）

商标异议·无效申请推理引擎（SJ-IRAC）：基于法条要件、证据链与风险分级的专业级审查与攻防系统。

Security Scan

This is an instruction-only, examiner-oriented trademark opposition/invalidation reasoning engine whose declared purpose, inputs, and runtime instructions are internally consistent and request no extra credentials, installs, or system access.

Purpose & Capability

The name, description, and the SKILL.md all describe a CNIPA-focused IRAC + evidence-chain engine. The required inputs (registry data, timeline, evidence inventory) match the stated legal analysis purpose; there are no requests for unrelated credentials, binaries, or external services.

Instruction Scope

All runtime instructions in SKILL.md and the included pipeline files are limited to intake of case materials, statutory mapping, evidence scoring, risk gates, and producing examiner-ready outputs. There are no instructions to read arbitrary system files, access environment variables, contact unknown endpoints, or exfiltrate data.

Install Mechanism

This is an instruction-only skill with no install spec and no code files to execute. That minimizes on-disk persistence and reduces supply-chain risk.

Credentials

The skill requires no environment variables, credentials, or config paths. The data it needs (case packet, registry numbers, evidence inventory) is proportional and explicitly listed in SKILL.md.

Persistence & Privilege

Skill does not request always:true and is user-invocable only. It does not attempt to modify other skills or system settings; autonomous invocation is allowed by default but is not coupled with elevated privileges or broad access.

Assessment

This skill appears coherent and low-risk because it is instruction-only and asks only for case materials. Before using it: (1) avoid submitting sensitive personal data unless necessary—redact or anonymize where possible; (2) treat outputs as advisory legal analysis and have a qualified attorney verify any submission; (3) confirm that citations and statutory references match your jurisdiction/current rules (the skill targets PRC/CNIPA); (4) remember the agent may run autonomously by default—only enable automated runs if you trust the agent's overall behavior and monitoring; and (5) if you need offline handling of highly sensitive exhibits (signed affidavits, originals), keep those out of automated inputs and manage them through your normal secure legal workflow.

商标异议·无效申请推理引擎（SJ-IRAC）

Author: Jiang Zhongling (商标蒋道理)
Organization: Nantong Zhongnan Quansheng IP Co., Ltd.
Version: 1.1.2 Last Updated: 2026-02-03

---

Summary (for ClawHub)

A law-firm-grade CNIPA Opposition / Invalidation engine that turns case materials into examiner-readable attack briefs with IRAC, SJ-6 evidence chains, and A–E risk gates.
No templates. No fluff. No fabricated facts.

---

What This Skill Does

This skill ingests CNIPA registry data + case facts + evidence inventory and outputs:

• Ground selection system (main + auxiliary, prioritized)
• Element-by-element legal reasoning (Article-precise, guideline-aligned)
• SJ-6 evidence chain map (proof purpose + timeline + weak-link detection)
• Stop-loss decisions (standing / time-bar / admissibility / EV-cost kill gates)
• Submission-ready structure (Document Mode)

Constraint: No generic AI writing. No speculative conclusions. Only verifiable, evidence-backed reasoning.

---

v1.1 Highlights (Operational Upgrades)

1. Standing & Limitation Control (Hard Gate)

• Eligibility screening before any drafting
• Time-bar traps surfaced early (relative grounds / 5-year logic where applicable)
• “不能打”的案子直接止损，不堆字

2. Procedural Control Review

• Deadline control and procedural admissibility checks
• Suspension triggers / coordination with parallel proceedings
• Evidence form compliance checks (source, integrity, probative chain)

3. Risk Engine Refinement (Kill Gates)

• Procedural / discretionary fatal-defect gates
• EV-cost stop-loss when expected value < cost
• Evidence weakness quantified and action-ranked

---

Scope & Positioning

Primary Scope

• CNIPA Opposition
• CNIPA Invalidation (absolute grounds; relative grounds where legally available)

Core Mission

Convert dispute materials into a decision-grade argument system:

• Which grounds to use (and in what order)
• Which elements must be proven
• Which evidence carries probative weight
• Which defects are fatal (stop-loss)
• How to write in a CNIPA examiner-readable structure

Not a Template Pack

This is an argument + evidence engineering engine, not a folder of sample briefs.

---

Legal Basis (Bounded Sources)

Operates strictly within:

• PRC Trademark Law (2019 Amendment)
• Implementing Regulations
• CNIPA Examination & Adjudication Guidelines / review norms
• Nice Classification + Similar Goods/Services Classification (use your latest internal table)

Prohibited

• Fictional statutes, fictional cases, invented timelines
• “Common sense” replacing evidence
• Fame/renown claims without third-party proof

---

Core Framework

1) IRAC (Mandatory, Examiner-Oriented)

1. Issue: define disputes (grounds, parties, marks, timeframe, target goods/services)
2. Rule: map statutes + guideline purpose + elements + burden/standard
3. Application: match evidence to elements (逐要件对应，不做假设)
4. Conclusion: enforceable outcome + next-step plan (补证/改路/止损)

2) SJ-6 Evidence Chain (Mandatory)

Each item is scored under:

1. Authenticity
2. Relevance
3. Completeness
4. Temporal validity
5. Logical consistency
6. Cross-examination resistance

Evidence organization rules

• Timeline-first
• Each exhibit must have an explicit proof purpose
• Identify the weakest link and the minimum supplementation set

3) Risk Module (A–E + Kill Gates)

Outputs include:

• Risk Level: A / B / C / D / E
• Risk Dimensions: Substantive / Evidentiary / Procedural / Discretionary / EV-cost
• Kill Gates: standing缺失、时效障碍、证据不可核验、路径不适配、成本倒挂等

---

Supported Scenarios

• Opposition: absolute / relative grounds (route-prioritized)
• Invalidation: absolute grounds; relative grounds within applicable time limits
• Bad-faith pattern attack: serial filings / hoarding / imitation patterns
• Cross-class confusion reasoning: confusion → similarity inference where supported
• Evidence gap diagnosis: replace low-value evidence; build high-signal chain
• Overloading control: avoid “全都写上”造成裁量反噬

---

Input Requirements (Minimum Viable Case Packet)

Provide at least:

1. Target trademark number(s), status, filing/registration dates
2. Parties and relationship clues (if any)
3. Designated goods/services + class(es)
4. Case timeline (publication/registration + prior use milestones)
5. Intended grounds (optional; engine can propose)
6. Evidence inventory: source / date / type / brief / proof purpose (if known)

If inputs are incomplete → conservative output by design.

---

Output Modes

Quick Mode (Fast Triage)

• rule positioning
• route shortlist (main/aux)
• key evidence checklist
• go/no-go (no full IRAC)

Pro Mode (IRAC + SJ-6 + Risk)

• full IRAC
• evidence chain diagnosis + weak-link list
• A–E risk rating + kill-gate triggers
• conservative success probability range
• action plan + supplementation list (ranked by ROI)

Document Mode (Submission-Ready)

• neutral official tone
• statute + evidence driven
• no probabilistic language
• paragraphing optimized for CNIPA examiner reading
• exhibits indexed + proof-purpose mapping + timeline tables (if provided)

---

Compliance & Hard Constraints

• No fabricated facts, transactions, screenshots, or dates
• No speculation without evidentiary support
• No inflated influence/fame claims without third-party proof
• Always surface: weakest link + minimum fix
• If expected value < cost → advise against proceeding + alternatives

---

Typical Use Cases (Law-Firm Grade)

• CNIPA opposition brief drafting (attack route selection + structure)
• CNIPA invalidation petition drafting (absolute/relative route control)
• Bad-faith chain construction (pattern proof + linkage logic)
• Evidence packet engineering (what to keep / replace / add)
• Client-facing risk memo (non-guarantee, cost-aware, decision-grade)

---

How to Use

1. Provide registry data + facts + evidence inventory
2. Choose mode: Quick / Pro / Document
3. Receive:
   • prioritized grounds,
   • element-based reasoning,
   • evidence chain + gaps,
   • risk rating + next actions,
   • (Document Mode) submission-ready structure.

---

Versioning Notes

• Patch (x.y.z): doc/consistency fixes
• Minor (x.y.0): new modules / workflow upgrades
• Major (x.0.0): architecture changes