prism_auto_config_map

棱镜车智-专利论文数据穿透

Use When

Use this skill when the user asks for automotive competitor configuration analysis, future vehicle planning, or feature timing decisions where Patsnap patent/literature/supply-chain evidence should explain the "why now" behind the recommendation.

Typical requests:

• "20-30 万 B 级 EREV SUV 哪些配置已经成为必备,哪些还有差异化机会"

• "对比理想 L6/L7、问界 M7/M8、零跑 C16、阿维塔 07、智界 R7,输出下一代 B-SUV 配置机会地图"

• "年轻家庭 25 万级 B-SUV,哪些配置该跟进,哪些适合作为专利壁垒型差异化"

• "电子外后视镜、地暖、二排腿托、座椅抽屉,哪些未来 6-18 个月值得布局"

• "把特斯拉、比亚迪、华为系作为头部标尺,看下一代智能座舱/舒适配置机会"

Do not expand into broad intelligence work. If the user asks for company strategy, financing, organization, FTO, or precise BOM/TCO, narrow the task back to patent/literature/supply-chain evidence for next-generation vehicle configuration decisions.

Core Audience

The report must include a visible "数据聚焦与服务对象" section. State that the analysis is designed for:

• 产品规划: define next-generation feature packages and trim logic.

• 战略/竞品情报: compare market leaders and detect timing windows.

• 研发预研: identify technical routes, patents, literature signals, and design constraints.

• 采购/供应链: find OEM/Tier0/Tier1/Tier2 linkage clues and sourcing risk.

• 售前/管理层 Demo: communicate the decision headline and evidence chain quickly.

Freshness And Forecast Rules

This skill is forward-looking. Always use the latest data available in the execution environment.

• Show 数据统计截止日期, Patsnap 数据截止日期, and 外部网页获取日期 in the hero area.

• Prefer data collected or refreshed as close as possible to the report date.

• If a source is stale, show a "前瞻性影响说明" card and lower confidence.

• Patent publication lag must be disclosed. Do not treat recent absence of patents as absence of activity.

• For 1-2 year forecasts, combine market fitment, Patsnap patent/literature trend, and supply-chain maturity.

Core Workflow

1. Define the competitor pool : price band, segment, powertrain, launch window, target persona, model list, and head-company benchmark.

2. Add head-company references : include or explicitly discuss Tesla, BYD, Huawei/Harmony Intelligent Mobility, and relevant China NEV or overseas leaders when the segment allows it. If one is excluded, state why.

3. Resolve product-domain hierarchy : if the prompt locks a domain such as intelligent chassis, smart cockpit, thermal management, battery/e-drive, seat comfort, or ADAS, first map product domain → level-1 technologies → level-2 configurations/components → analysis objects.

4. Run configuration benchmarking : compare seat, interior, exterior, cabin, ADAS/comfort, thermal, battery, and safety-related configurations. Keep the user's model pool; use head companies only as benchmarks.

5. Build QFD / House of Quality : translate graded pain points into customer needs, engineering characteristics, competitor satisfaction, Patsnap evidence strength, target values, and planning actions.

6. Run Matrix Analysis : combine penetration, QFD weight, lifecycle stage, patent barrier, supply-chain maturity, and head-company benchmark into must_follow / differentiation_bet / watch / avoid_or_design_around.

7. Apply Patsnap patent-paper dashboard + technology penetration wall : for every level-1 technology in the product-domain map, build a patent/literature pool, then show market phenomenon → annual patent/literature trend → key applicants → co-applicants/inventor signals → core patent highlights → problem-solution-effect → supply-chain feasibility → 1-2 year forecast.

8. Apply supply-chain linkage lens : map configuration → function module → key components → core technology → OEM/Tier0/Tier1/Tier2 → China/overseas region → cooperation signal → evidence source.

9. Build the one-line decision : create headline_conclusion, one concise sentence explaining the future opportunity, risk, or action.

10. Render HTML : use assets/report_template.html and local assets under assets/vehicle_visuals/. Image generation or web image fetching is optional enhancement, never a hard dependency.

Required Theory Tags

Every major conclusion must carry at least one theory tag. The final methodology display should focus on QFD / House of Quality and Matrix Analysis; other methods are supporting tags.

• [QFD / House of Quality]

• [Matrix Analysis]

• [Benchmarking]

• [Kano]

• [S-Curve]

• [Patent Barrier]

• [Patsnap Penetration]

• [Supply Linkage]

The final report must include a visible "理论依据 / Methodology" section so non-specialists can see why the analysis is grounded.

Evidence Rules

• For market facts, prefer official or high-credibility sources. Read references/data_sources.md before making source claims.

• For Patsnap-driven claims, read references/patsnap_data_lens.md and show the full penetration chain in the main report, not only in the appendix.

• For supplier or supply-chain claims, read references/supply_chain_lens.md; supplier findings must be source-backed or explicitly marked as "待验证供应商线索".

• Do not use patent volume as a proxy for sales, user demand, or actual take rate.

• A "differentiation bet" requires at least one Patsnap strong signal: patent trend, literature foresight, Tier1/Tier2 supply-chain evidence, or a key patent with clear design intent.

• A "must follow" requires market or user evidence: high penetration, official fitment, clear VOC demand, or safety/regulatory basis.

• A "design-around/avoid" must state whether the reason is patent barrier, user rejection, regulatory uncertainty, cost, or implementation risk.

• Patent evidence for key configurations must be traceable to publication number, applicant, filing/publication date, legal status where available, tool/source, and data cutoff.

• OEM-supplier links must be labeled as one of: 已验证合作, 专利联合申请, 产品/公告线索, 待验证供应商线索.

• If a data source is unavailable, output a data gap card and lower confidence. Never invent source-backed numbers.

Head-Company Benchmark Rule

For vehicle benchmarking, do not miss domestic and overseas leaders:

• Default head-company benchmarks: Tesla, BYD, Huawei/Harmony Intelligent Mobility.

• Add Volkswagen, Toyota, Hyundai, Mercedes-Benz, or other overseas leaders when the product domain or market scope makes them relevant.

• Head companies are benchmarks, not replacements for the user's specified competitor pool.

• If Tesla, BYD, or Huawei/Harmony Intelligent Mobility is not applicable, explicitly state the exclusion reason in the report.

Product-Domain First Rule

When the user names or implies a product domain, the report must first show the domain map before judging configurations:

产品域 → 一级技术 → 二级配置/零部件 → 竞品配置 → 专利/文献/供应链证据 → 规划动作

Typical domains: smart cockpit, intelligent chassis, ADAS/safety, thermal management, battery/e-drive, seat and comfort, exterior and body electronics.

Patent-Paper Dashboard And Patent Pool Rules

• Do not output a raw patent list as the main analysis.

• Use a patent-paper dashboard by company, product domain, and level-1 technology: annual patent growth, literature/paper signal, applicant concentration, co-applicant network, inventor signal, core patent highlights, patent pool scale, and planning implications.

• Every level-1 technology identified in the product-domain map must have a corresponding technology penetration card. If the report analyzes six level-1 technologies, render six cards; do not show only selected examples.

• Each technology card must include product domain, level-1 technology, linked level-2 configurations/components, patent pool count, search window, query terms/synonyms, literature/paper count and signal, layout cluster distribution, representative patents, applicant/supplier role split, 1-2 year forecast, confidence, and data gaps.

• If literature/paper retrieval has no results, explicitly show "论文信号暂无，以专利为主" and continue. Do not fail report generation because literature data is unavailable.

• Treat the patent pool as the complete retrieved set for that level-1 technology. The main report should show summary metrics and 3-6 representative patents; the full patent and literature lists should live in a detail page or same-page appendix.

• Preferred static output structure: index.html for the main report and patent_pools/{technology_slug}.html for one detail page per level-1 technology. If the host agent can only produce one HTML file, degrade to same-page anchors such as #pool-{technology_slug} while preserving the same content structure.

• Patent pool detail pages must show the full patent pool, full literature/paper pool when available, layout clusters, representative patent explanations, source notes, data gaps, and links to Patsnap/source detail pages where available.

• Do not generate one HTML page per individual patent; that is too slow and unnecessary. Individual patents should link to Patsnap/source detail URLs.

• Core patent cards must include publication number, applicant, filing/publication date, legal status when available, technical highlight, product planning implication, detail URL when available, and LLM limitation note.

• Inventor signals may indicate technical continuity, team clustering, or repeated research direction, but must not be used alone to infer employment, commercial cooperation, or supplier relationship.

HTML Generation And Portability Rules

• Generated reports must be static HTML that can open offline and use relative paths for packaged assets.

• Do not depend on external scripts, CDN chart libraries, or remote images.

• Use assets/report_template.html as the main visual skeleton. Reuse its CSS for both index.html and patent pool detail pages.

• Use summary charts in the main report: sparklines, layout cluster bars, applicant role cards, and heatmaps. Put long tables and full pools into detail pages or same-page appendices.

• The template must stretch with data volume: rich datasets should not be truncated; sparse datasets should still render with data gap cards and confidence reduction.

• Ensure responsive behavior: desktop can use two-column or card-grid layouts; mobile should fall back to single column with horizontally scrollable tables and wrapped long text.

Delivery-Grade Report Rules

When HTML output is requested, generate a delivery-grade multi-page report by default, not a bare workflow validation page.

• Hero right side must contain: main visual, caption/status tags, four methodology cards (QFD, Matrix, Patent, Lifecycle), and three or more test/report summary metrics.

• Every core section must include at least one visual expression: local image, static CSS/SVG chart, relationship card, heatmap, evidence card, or compact dashboard. Do not leave long sections as plain prose only.

• Product-domain technology map must be rendered as a table with: product domain, level-1 technology, level-2 configurations/components, related models/benchmarks, patent/literature pool, supplier role, and planning action.

• Configuration penetration heatmap must include a data-gap column distinguishing official confirmation, public-page confirmation, unverified, and authorization-required configuration-library confirmation.

• Main Patsnap section must include 3-5 representative patent PSE cards per level-1 technology. PSE means technical problem, solution, expected effect, product planning implication, and risk/design-around advice.

• Add an "AI 前瞻创新发现" module after the patent-paper penetration wall. Recommend 3-5 forward-looking configuration ideas based on patent pools, literature signals, and target scenarios. Each idea must state related technology domain, discovery basis, vehicle-fit concept, 1-2 year feasibility, supply-chain maturity, risk, and recommended action.

• Patent pool detail pages must include a local visual block, retrieval scope, layout clusters, representative patents with detail URLs, full patent list, full literature list, applicant/supplier view, and data gap notes.

• Each patent row should link to Patsnap/source details when available. Each paper/literature row should link to a source URL when available; otherwise show "详情链接待补".

Image Stability And Packaging Rules

Final report images must remain visible after the report is moved or zipped.

• Copy assets/vehicle_visuals/ into the report output directory before final delivery.

• Main report image paths must use assets/vehicle_visuals/....

• Patent pool detail page image paths must use ../assets/vehicle_visuals/....

• data-fallback paths must follow the same relative-path rule as src.

• Do not use /Users/..., /tmp/..., ../prism_auto_config_map/..., or remote image URLs as final image paths.

• Photoreal assets and SVG fallback assets are packaged visuals only; they are not evidence and must not be described as official OEM images.

Real Test Output Rules

For real skill tests and demo reruns:

• Create a new output directory for each test run. Do not overwrite previous reports unless the user explicitly asks.

• Prefer index.html plus patent_pools/*.html plus assets/ plus a zip package.

• Name delivery test packages clearly, for example prism_auto_config_map_0529_delivery_upgraded.zip for a skill package or {scenario}_report.zip for a generated report.

• Before delivery, validate HTML parsing, local multi-page links, image path existence, zip contents, and offline portability after unzip.

Technology Lifecycle Strategy

Classify each key configuration or technology as:

• 引领技术: benchmark-setting, high-end signal, usually lower penetration but strong strategic value.

• 上升技术: rising adoption, meaningful user value, suitable as selling point.

• 成熟技术: stable supply and broad fitment, suitable as baseline.

• 接近退出技术: useful for cost control but with experience or brand risk.

• 退出技术: should not appear in high-end or next-generation flagship recommendations.

Pricing guidance:

• 高端产品: must include leading technology as benchmark, rising technology as selling point, mature technology as base, and no exiting technology.

• 中端产品: use rising + mature technologies to build value-for-money competitiveness.

• 低端产品: use mature + near-exit technologies for cost control, with explicit experience and brand-risk disclosure.

Output Structure

Default output is a local HTML report using assets/report_template.html. The report must contain these sections:

1. Hero decision summary with headline_conclusion

2. Data freshness and forward-looking caveat

3. 数据聚焦与服务对象

4. Methodology focus: QFD / House of Quality + Matrix Analysis, with supporting tags summarized

5. 产品域技术地图

6. Competitor pool and head-company benchmark

7. Configuration penetration heatmap

8. QFD / House of Quality with pain-point levels

9. Matrix Analysis opportunity board

10. Patsnap patent-paper dashboard + technology penetration wall, including patent pools for every level-1 technology

11. OEM-supplier-patent relationship map, including overseas supply-chain leads where evidence supports them

12. Technology lifecycle x price-band strategy matrix

13. Next-generation configuration recommendation + evidence appendix

When responding in chat, provide a concise summary and the generated HTML path. If HTML generation is not requested or impossible, still structure the markdown in the same order.

If multi-page HTML generation is supported, produce:

• index.html: main decision report with summary cards for every level-1 technology.

• patent_pools/{technology_slug}.html: one patent/literature pool detail page per level-1 technology.

If multi-page generation is not supported, keep a single index.html and render the same detail content under same-page anchors.

Visual Asset Rules

• Prefer local photoreal assets from assets/vehicle_visuals/photo_real/.

• Use local SVG assets under assets/vehicle_visuals/ as low-fidelity fallback only.

• Read assets/vehicle_visuals/metadata.json before selecting a visual.

• Match by brand, market, powertrain, or scene. If no exact match exists, use the closest generic scene.

• If no image asset is available, render CSS/chart placeholders and continue.

• If image generation or web image fetching is available, it may be used as an enhancement only. Record source type and usage notes; never fail the skill because image generation is unavailable.

• Do not claim local AI-generated photoreal or SVG visuals are official OEM images.

• Photoreal assets are visual enhancement only; never use them as analytical evidence.

Reference Loading

• Read references/methodology.md when defining the analysis lens or explaining theory basis.

• Read references/data_sources.md before collecting or ranking non-Patsnap evidence.

• Read references/patsnap_data_lens.md before making patent, literature, S-curve, or applicant claims.

• Read references/supply_chain_lens.md before naming suppliers, Tier0/Tier1/Tier2 roles, or OEM-supplier relationships.

• Read references/output_templates.md before drafting the final report.

• Read references/config_taxonomy.md when normalizing configuration names.

• Read references/demo_prompts.md for stable demo scenarios and acceptance checks.

Quality Gate

Before final delivery, verify:

• The report contains headline_conclusion in a prominent hero position.

• The report shows data cutoff dates and any forward-looking limitation.

• The report contains "数据聚焦与服务对象" and "理论依据 / Methodology".

• Key conclusions include theory tags.

• Methodology display focuses on QFD / House of Quality and Matrix Analysis; other methods are summarized as supporting tags.

• If a product domain is named, the report contains product-domain technology mapping before detailed configuration analysis.

• Each differentiation bet has a Patsnap evidence row.

• Each named supplier or supplier category has evidence, maturity, role labels, and relationship status; otherwise it is marked as a lead to verify.

• Head-company benchmark includes Tesla, BYD, Huawei/Harmony Intelligent Mobility, or a stated reason for exclusion; overseas leaders are added when relevant.

• Patent and literature/paper evidence is displayed as dashboard/cards and technology-level patent pools, not as an unstructured raw list in the main report.

• Every level-1 technology in the product-domain map has a patent/literature pool summary card and a detail page or same-page anchor.

• Literature/paper absence is explicitly marked as "论文信号暂无，以专利为主" and does not block report generation.

• QFD pain points are graded by severity/importance and linked to engineering characteristics.

• Inventor signals are used only as technical continuity clues with limitations.

• Key configurations include 1-2 year future-state labels for technology, OEM/Tier0, and price band.

• Technology lifecycle strategy is visible for high-end / mid-end / low-end product positioning.

• Non-Patsnap data sources have reliability grade and acquisition path.

• The HTML can open offline and does not depend on external scripts.

• Missing image generation or web access does not block report generation.

使用前配置

本 Skill 依赖智慧芽开放平台 MCP 服务：

• 完成安装、初次使用时需进行自检，参见 README.md

• 用户需完成账号授权，并确保 Agent 环境已启用对应 MCP 工具

• 若未完成配置，本 Skill 只能提供分析框架，无法检索实时数据或生成基于数据库的结论

• 缺少MCP配置时，引导用户参照 README.md 在 [ open.zhihuiya.com ]( https://open.zhihuiya.com/ ) 获取MCP。