R&D Initiation Review
Provided by Patsnap Eureka.
Use this skill when the decision object is a concrete project, proposal,
innovation package, or initiation document set. The goal is to judge whether
it should move forward, how similar it is to public prior work, which
innovation claims still stand, and what evidence or materials are still missing.
The customer-visible value should land first on four questions:
- Is this project worth advancing?
- How close is it to public prior work?
- Do the innovation claims hold up?
- What is still missing before the next gate?
Use When
- The user provides a project or proposal and asks whether it is worth advancing
- Formal initiation review report for a concrete project
- Pre-screening / public novelty pre-check / client-facing pre-screen
- Go/no-go or budget-release recommendation around a proposal
- Project-level novelty / feasibility / trend / risk review
- Innovation-point-by-innovation-point assessment
- Management-ready review bundle with evidence pack
Do Not Use
- Pure technology-route comparison with no concrete project object → route to
tech-route-comparison
- Company-by-topic technical profile or player landscape → route to
company-tech-profile or competitive-landscape
- Legal patent opinion, FTO, infringement, or litigation analysis
- Drafting the original proposal from scratch (this skill reviews, not writes)
- Generic company overview with no project-level review depth
Review Modes
| Mode |
When To Use |
Typical Output |
screen |
Pre-screening, early-stage review, customer-facing presales |
Quick go/no-go + novelty boundary + key gaps (6-8 pages) |
review |
Standard initiation review |
Full review report + novelty note + evidence pack (10-14 pages) |
innovation |
Innovation-point deep review |
Point-by-point innovation assessment + differentiation table (8-12 pages) |
assurance |
Expert review, committee-ready package |
Deep evidence-sufficiency audit + rating logic + stage-gate matrix (12-18 pages) |
Default mode selection:
- If the request is for pre-screening, novelty pre-check, overlap check,
"is it worth advancing", or an early customer-facing review → default to
screen
- If the request is for expert review, go/no-go, budget release, evidence
sufficiency, or committee-ready bundle → default to
assurance
- Otherwise → default to
review
Core Principles
Proposal As Input, Not Truth
The proposal is the starting point for extraction, not a source of verified
facts. A proposal claim does not become a review conclusion unless it is either
corroborated externally or explicitly labeled as proposal-only.
Four-Layer Evidence Separation
Every substantive claim in the review must be tagged as one of:
- Proposal-stated: what the proposal says (quoted or summarized)
- Externally corroborated fact: confirmed by patent, paper, or official source
- Evidence-backed inference: reasonable conclusion from multiple signals
- Open gap: insufficient evidence — state what is missing and what would resolve it
Do not let proposal-stated claims silently become review conclusions.
Novelty As First-Class Visible Module
Novelty search is not a side effect of baseline retrieval. The review must
clearly state:
- Novelty-search scope (databases, time window, field scope)
- Main comparison objects (prior art, prior routes, existing solutions)
- Point-by-point overlap assessment
- Residual differentiators
- Explicit novelty judgment level for each major claim
- Search limitations and unsearched areas
When novelty is downgraded because of prior art, build a structured
differentiation-boundary table showing overlap, residual differentiators,
and replacement risk.
Material Completeness As First-Class Module
Even when the platform cannot solve the client's full data governance problem,
the report should clearly say which missing materials or data are currently
blocking:
- novelty or overlap boundary judgment
- feasibility judgment
- next-step initiation actions
If internal archives or non-public project databases are not available, say so
explicitly. Do not imply that a public-only retrieval pass proves exhaustive
internal duplicate-project exclusion.
Tool Routing And Fallback
This skill works across multiple tool environments. Before retrieval, detect
which capabilities are available and select the highest tier that is reachable.
Tier 0 (Always First): Proposal Materials
- User-provided proposal, initiation report, innovation package
- Extract project summary, research theme, innovation points, promised outputs,
and proposal-stated claims BEFORE any external search
- Evidence grade: proposal-stated (not externally verified)
Tier 1 (Recommended): Structured Patent/Paper Retrieval
- Use the host's best structured patent and paper retrieval stack to test
novelty, feasibility, and frontier evidence around normalized innovation points.
- This usually means structured search plus record-level deep fetch.
- Evidence grade: S/A
Tier 2 (Fallback): Exa Search + Companion Skills
- Use the host's best broad web research tool plus scholarly or finance
companion lanes when available.
- Exa, Tavily, Brave, official filings, and domain-specific research databases
are good examples, not hard requirements.
- Evidence grade: A/B
- Coverage loss vs Tier 1: no structured patent field search, no assignee
filtering, novelty conclusions less precise
Tier 3 (Fallback): Generic Web Tools
- Use generic web search and page/PDF reading tools available in the host.
- Evidence grade: B/C
Tier 4 (Minimum): Pure LLM + Proposal Materials
- No external tools required
- Evidence grade: C/U
- The review can still extract and organize proposal claims, identify logical
gaps, and flag what external evidence would be needed
Routing Rules
- Always start from proposal materials (Tier 0) regardless of tool availability.
- Select the highest available external tier for verification.
- When a tier is unavailable, explicitly state the downgrade.
- Do not pretend a public-only pass equals full duplicate-project exclusion.
- When using Tier 3 or 4, increase explicit gap statements and reduce
confidence on novelty and completeness claims.
Minimum Working Files
Create or update these files in a writable run folder:
request.md
workplan.md
method_decisions.md
query_log.csv
source_index.csv
claim_ledger.csv
report.md
novelty-note.md
Recommended subfolders are described in references/workflow.md.
Default Workflow
Step 0: Freeze Scope
Confirm or infer the following before any retrieval:
project_name: the concrete project or proposal object
topic: the research theme (infer from materials if not stated)
decision_gate: pre-approval / go-no-go / budget release / expert review
mode: screen / review / innovation / assurance
deliverable: brief / report / docx / pdf / evidence-pack
audience: client / internal / committee / management
time_window: default last 3-5 years
review_focus: novelty / feasibility / trend / risk / rating (can be multiple)
innovation_scope: extract from materials or user-specified innovation points
If the user supplied materials but did not name the research theme, infer a
provisional theme from the materials and log that it is provisional.
If the user did not provide a concrete project object at all, do not force
this skill — redirect to tech-route-comparison or company-tech-profile.
Step 1: Extract Proposal Materials
Before any external search:
- Read the proposal and extract:
- Project summary and research theme
- Stated innovation points (normalize into a short comparison set)
- Promised outputs and deliverables
- Proposal-stated claims about novelty, feasibility, and market
- Technical path description (main chain, dependencies, measurement approach)
- Build an innovation register: each innovation point as an atomic item with
the proposal's own language preserved.
- Identify what the proposal does NOT address: missing data, unstated
assumptions, logical gaps.
- Tag everything as
proposal-stated at this stage.
Step 2: Retrieve External Evidence
- Search for prior art and prior routes around each normalized innovation point.
- Tier 1: structured patent/paper search around each innovation point
- Tier 2: web research plus scholarly companion search
- Tier 3/4: generic web or proposal-only reasoning
- For each innovation point, run a dedicated novelty check:
- Search scope: patents + papers + standards in the topic area
- Record: database/source scope, time filter, field scope, linked novelty
point, screening rule
- Identify: closest prior art, overlap degree, residual differentiators
- Search for feasibility and frontier support:
- Are the proposed technical approaches validated elsewhere?
- What is the current state of the art?
- Are there known failure modes or blockers?
- Search for trend and policy context when relevant:
- Industry direction, regulatory environment, standards status
- Only use web sources for current-status claims (standards, regulation,
launch status, productization proof)
Step 3: Build Analysis By Mode
Select modules based on the current mode:
| Module |
screen |
review |
innovation |
assurance |
| Proposal extraction |
✓ |
✓ |
✓ |
✓ |
| Novelty boundary |
✓ |
✓ |
✓ |
✓ |
| Innovation-point assessment |
light |
standard |
deep |
deep |
| Feasibility check |
light |
standard |
light |
deep |
| Trend/frontier context |
optional |
standard |
optional |
standard |
| Material completeness |
✓ |
✓ |
✓ |
✓ |
| Scoring/rating |
skip |
optional |
optional |
required |
| Stage-gate matrix |
skip |
skip |
skip |
required |
| Counterevidence pass |
skip |
optional |
optional |
required |
| Governance review |
skip |
skip |
skip |
optional |
Step 4: Synthesize And Draft
Write the report following the mode-appropriate output skeleton.
For customer-facing or presales deliverables, put the visible value on the
front pages in this order:
- Recommendation
- Novelty and overlap boundary
- Key issues and risks
- Missing materials, data, and next steps
Do not let deep technical decomposition crowd out these questions in the
main narrative.
Output Skeletons
Screen Mode
- Executive recommendation (proceed / proceed with conditions / hold)
- Novelty boundary summary (1-2 paragraphs)
- Key risks and blockers
- Missing materials and next steps
Review Mode
- Executive recommendation
- Project scope and proposal-stated claims
- Novelty and overlap judgment (with comparison table)
- Feasibility and evidence sufficiency
- Main risks and material gaps
- Recommendation and conditions
Innovation Mode
- Executive summary of innovation assessment
- Innovation-point-by-point analysis (each point: proposal claim → external
evidence → overlap → residual differentiator → judgment)
- Differentiation boundary table
- Risks to innovation claims
- Recommendation
Assurance Mode
- Executive recommendation with confidence level
- Project scope and proposal coherence assessment
- Novelty and overlap (with detailed comparison attachment)
- Feasibility, evidence sufficiency, and technical decomposition
- Scoring and rating logic (with explicit rubric)
- Stage-gate matrix (conditions, owners, timelines, non-fulfillment consequences)
- Material gaps and data readiness
- Counterevidence and weakening conditions
Every claim must cite its source type and identifier, e.g., [Patent: CN1234567B],
[Paper: DOI or title], [Web: source name], [Proposal: section/page].
Formal Output Rules
For formal review audiences such as large enterprises, public-sector bodies,
and research institutions:
- Remove empty opening filler such as trend-setting or era-setting lead-ins.
- Remove promotional language and replace it with concrete review judgments.
- Replace vague attribution with named evidence and explicit citations.
- Rewrite formulaic paired constructions into direct standalone judgments.
- Do not leak control-plane jargon such as
proposal-only, prior-art,
assurance, gate, or owner into the main report unless the term is
necessary and clearly explained.
Headings and body text should use plain formal wording suitable for a written
review or committee memo.
When novelty is a live decision question, promote the method and point-by-point
comparison into novelty-note.md. A suitable file title is
Public Novelty and Prior-Art Review Note or Novelty and Benchmarking Appendix.
Completion Gates
All Modes
Screen Mode Additional
Review Mode Additional
Innovation Mode Additional
Assurance Mode Additional
Guardrails
- The proposal is an input object, not a source of truth. Do not let proposal
rhetoric substitute for external evidence.
- Do not imply that a public-only retrieval pass proves exhaustive coverage or internal
duplicate-project exclusion.
- Do not let raw search hit counts or internal ledger jargon become the main
management-facing argument.
- Treat proposal promises, roadmap claims, and commercialization statements as
signals until corroborated.
- Keep scoring as a support layer, not the main selling point. Prefer direct
judgments (proceed / proceed with conditions / hold) over dense weighted scoring tables
in the main body.
- When novelty is downgraded, do not stop at prose caution — build a structured
differentiation-boundary table.
- Do not force a strong go/no-go when the evidence base is thin — prefer
conditional recommendations with explicit next-evidence-gathering steps.
- Separate the novelty methodology (scope, queries, limitations) from the
novelty conclusion.
- Do not mix proposal-stated claims with externally verified facts without
clear labeling.
- For formal reports, do not leak control-plane jargon into the
management-facing narrative.
- Treat internal duplicate-project exclusion as data-dependent — do not promise
it when internal project history is unavailable.
Load These Files Only As Needed
- references/method-benchmark.md
- references/domain-playbooks.md
- references/workflow.md
- references/source-routing.md
- references/deliverables.md
- references/evidence-schema.md
- references/quality-gates.md
- templates/request-template.md
- templates/workplan-template.md
- templates/report-outline.md
1---2copyright: "Copyright © Patsnap. All rights reserved."3name: rd-initiation-review4description:5 R&D project initiation pre-screen and proposal audit for go/no-go decisions,6 public novelty boundary review, innovation-point assessment, and evidence-backed7 project rating. Use when the user asks for project initiation pre-screening,8 initiation review, proposal review, R&D project evaluation, proposal-package9 review, novelty pre-screening, innovation-point review, project rating, or10 wants a formal review around a concrete project, proposal, or research-package11 material set — even if they only provide the proposal and do not explicitly12 say "review".13argument-hint: "[project proposal, initiation report, or review request]"14provider: "Patsnap Eureka"15compatibility: "Designed for Claude Code, Codex, and similar agent runtimes that can read/write local files and call whatever search tools are available."16deliverable-default: "Structured Markdown review report plus novelty note and traceable evidence files; docx/pdf export is optional."17fallback-policy: "Start from proposal materials; prefer structured patent/paper retrieval when available; otherwise downgrade through domain-specific sources, Exa, Tavily, Brave, web, and a known-URL reader without blocking the run."18metadata:19 source: https://skillmd.com/skills/patsnap/rd-initiation-review20---2122# R&D Initiation Review2324Provided by Patsnap Eureka.2526Use this skill when the decision object is a concrete project, proposal,27innovation package, or initiation document set. The goal is to judge whether28it should move forward, how similar it is to public prior work, which29innovation claims still stand, and what evidence or materials are still missing.3031The customer-visible value should land first on four questions:32331. Is this project worth advancing?342. How close is it to public prior work?353. Do the innovation claims hold up?364. What is still missing before the next gate?3738## Use When3940- The user provides a project or proposal and asks whether it is worth advancing41- Formal initiation review report for a concrete project42- Pre-screening / public novelty pre-check / client-facing pre-screen43- Go/no-go or budget-release recommendation around a proposal44- Project-level novelty / feasibility / trend / risk review45- Innovation-point-by-innovation-point assessment46- Management-ready review bundle with evidence pack4748## Do Not Use4950- Pure technology-route comparison with no concrete project object → route to51 `tech-route-comparison`52- Company-by-topic technical profile or player landscape → route to53 `company-tech-profile` or `competitive-landscape`54- Legal patent opinion, FTO, infringement, or litigation analysis55- Drafting the original proposal from scratch (this skill reviews, not writes)56- Generic company overview with no project-level review depth5758## Review Modes5960| Mode | When To Use | Typical Output |61|------|------------|----------------|62| `screen` | Pre-screening, early-stage review, customer-facing presales | Quick go/no-go + novelty boundary + key gaps (6-8 pages) |63| `review` | Standard initiation review | Full review report + novelty note + evidence pack (10-14 pages) |64| `innovation` | Innovation-point deep review | Point-by-point innovation assessment + differentiation table (8-12 pages) |65| `assurance` | Expert review, committee-ready package | Deep evidence-sufficiency audit + rating logic + stage-gate matrix (12-18 pages) |6667Default mode selection:6869- If the request is for pre-screening, novelty pre-check, overlap check,70 "is it worth advancing", or an early customer-facing review → default to `screen`71- If the request is for expert review, go/no-go, budget release, evidence72 sufficiency, or committee-ready bundle → default to `assurance`73- Otherwise → default to `review`7475## Core Principles7677### Proposal As Input, Not Truth7879The proposal is the starting point for extraction, not a source of verified80facts. A proposal claim does not become a review conclusion unless it is either81corroborated externally or explicitly labeled as proposal-only.8283### Four-Layer Evidence Separation8485Every substantive claim in the review must be tagged as one of:86871. **Proposal-stated**: what the proposal says (quoted or summarized)882. **Externally corroborated fact**: confirmed by patent, paper, or official source893. **Evidence-backed inference**: reasonable conclusion from multiple signals904. **Open gap**: insufficient evidence — state what is missing and what would resolve it9192Do not let proposal-stated claims silently become review conclusions.9394### Novelty As First-Class Visible Module9596Novelty search is not a side effect of baseline retrieval. The review must97clearly state:9899- Novelty-search scope (databases, time window, field scope)100- Main comparison objects (prior art, prior routes, existing solutions)101- Point-by-point overlap assessment102- Residual differentiators103- Explicit novelty judgment level for each major claim104- Search limitations and unsearched areas105106When novelty is downgraded because of prior art, build a structured107differentiation-boundary table showing overlap, residual differentiators,108and replacement risk.109110### Material Completeness As First-Class Module111112Even when the platform cannot solve the client's full data governance problem,113the report should clearly say which missing materials or data are currently114blocking:115116- novelty or overlap boundary judgment117- feasibility judgment118- next-step initiation actions119120If internal archives or non-public project databases are not available, say so121explicitly. Do not imply that a public-only retrieval pass proves exhaustive122internal duplicate-project exclusion.123124## Tool Routing And Fallback125126This skill works across multiple tool environments. Before retrieval, detect127which capabilities are available and select the highest tier that is reachable.128129### Tier 0 (Always First): Proposal Materials130131- User-provided proposal, initiation report, innovation package132- Extract project summary, research theme, innovation points, promised outputs,133 and proposal-stated claims BEFORE any external search134- Evidence grade: proposal-stated (not externally verified)135136### Tier 1 (Recommended): Structured Patent/Paper Retrieval137138- Use the host's best structured patent and paper retrieval stack to test139 novelty, feasibility, and frontier evidence around normalized innovation points.140- This usually means structured search plus record-level deep fetch.141- Evidence grade: S/A142143### Tier 2 (Fallback): Exa Search + Companion Skills144145- Use the host's best broad web research tool plus scholarly or finance146 companion lanes when available.147- Exa, Tavily, Brave, official filings, and domain-specific research databases148 are good examples, not hard requirements.149- Evidence grade: A/B150- Coverage loss vs Tier 1: no structured patent field search, no assignee151 filtering, novelty conclusions less precise152153### Tier 3 (Fallback): Generic Web Tools154155- Use generic web search and page/PDF reading tools available in the host.156- Evidence grade: B/C157158### Tier 4 (Minimum): Pure LLM + Proposal Materials159160- No external tools required161- Evidence grade: C/U162- The review can still extract and organize proposal claims, identify logical163 gaps, and flag what external evidence would be needed164165### Routing Rules166167- Always start from proposal materials (Tier 0) regardless of tool availability.168- Select the highest available external tier for verification.169- When a tier is unavailable, explicitly state the downgrade.170- Do not pretend a public-only pass equals full duplicate-project exclusion.171- When using Tier 3 or 4, increase explicit gap statements and reduce172 confidence on novelty and completeness claims.173174## Minimum Working Files175176Create or update these files in a writable run folder:177178- `request.md`179- `workplan.md`180- `method_decisions.md`181- `query_log.csv`182- `source_index.csv`183- `claim_ledger.csv`184- `report.md`185- `novelty-note.md`186187Recommended subfolders are described in [references/workflow.md](references/workflow.md).188189## Default Workflow190191### Step 0: Freeze Scope192193Confirm or infer the following before any retrieval:194195- `project_name`: the concrete project or proposal object196- `topic`: the research theme (infer from materials if not stated)197- `decision_gate`: pre-approval / go-no-go / budget release / expert review198- `mode`: screen / review / innovation / assurance199- `deliverable`: brief / report / docx / pdf / evidence-pack200- `audience`: client / internal / committee / management201- `time_window`: default last 3-5 years202- `review_focus`: novelty / feasibility / trend / risk / rating (can be multiple)203- `innovation_scope`: extract from materials or user-specified innovation points204205If the user supplied materials but did not name the research theme, infer a206provisional theme from the materials and log that it is provisional.207208If the user did not provide a concrete project object at all, do not force209this skill — redirect to `tech-route-comparison` or `company-tech-profile`.210211### Step 1: Extract Proposal Materials212213Before any external search:2142151. Read the proposal and extract:216 - Project summary and research theme217 - Stated innovation points (normalize into a short comparison set)218 - Promised outputs and deliverables219 - Proposal-stated claims about novelty, feasibility, and market220 - Technical path description (main chain, dependencies, measurement approach)2212. Build an innovation register: each innovation point as an atomic item with222 the proposal's own language preserved.2233. Identify what the proposal does NOT address: missing data, unstated224 assumptions, logical gaps.2254. Tag everything as `proposal-stated` at this stage.226227### Step 2: Retrieve External Evidence2282291. Search for prior art and prior routes around each normalized innovation point.230 - Tier 1: structured patent/paper search around each innovation point231 - Tier 2: web research plus scholarly companion search232 - Tier 3/4: generic web or proposal-only reasoning2332. For each innovation point, run a dedicated novelty check:234 - Search scope: patents + papers + standards in the topic area235 - Record: database/source scope, time filter, field scope, linked novelty236 point, screening rule237 - Identify: closest prior art, overlap degree, residual differentiators2383. Search for feasibility and frontier support:239 - Are the proposed technical approaches validated elsewhere?240 - What is the current state of the art?241 - Are there known failure modes or blockers?2424. Search for trend and policy context when relevant:243 - Industry direction, regulatory environment, standards status244 - Only use web sources for current-status claims (standards, regulation,245 launch status, productization proof)246247### Step 3: Build Analysis By Mode248249Select modules based on the current mode:250251| Module | screen | review | innovation | assurance |252|--------|--------|--------|------------|-----------|253| Proposal extraction | ✓ | ✓ | ✓ | ✓ |254| Novelty boundary | ✓ | ✓ | ✓ | ✓ |255| Innovation-point assessment | light | standard | deep | deep |256| Feasibility check | light | standard | light | deep |257| Trend/frontier context | optional | standard | optional | standard |258| Material completeness | ✓ | ✓ | ✓ | ✓ |259| Scoring/rating | skip | optional | optional | required |260| Stage-gate matrix | skip | skip | skip | required |261| Counterevidence pass | skip | optional | optional | required |262| Governance review | skip | skip | skip | optional |263264### Step 4: Synthesize And Draft265266Write the report following the mode-appropriate output skeleton.267268For customer-facing or presales deliverables, put the visible value on the269front pages in this order:2702711. Recommendation2722. Novelty and overlap boundary2733. Key issues and risks2744. Missing materials, data, and next steps275276Do not let deep technical decomposition crowd out these questions in the277main narrative.278279## Output Skeletons280281### Screen Mode2822831. Executive recommendation (proceed / proceed with conditions / hold)2842. Novelty boundary summary (1-2 paragraphs)2853. Key risks and blockers2864. Missing materials and next steps287288### Review Mode2892901. Executive recommendation2912. Project scope and proposal-stated claims2923. Novelty and overlap judgment (with comparison table)2934. Feasibility and evidence sufficiency2945. Main risks and material gaps2956. Recommendation and conditions296297### Innovation Mode2982991. Executive summary of innovation assessment3002. Innovation-point-by-point analysis (each point: proposal claim → external301 evidence → overlap → residual differentiator → judgment)3023. Differentiation boundary table3034. Risks to innovation claims3045. Recommendation305306### Assurance Mode3073081. Executive recommendation with confidence level3092. Project scope and proposal coherence assessment3103. Novelty and overlap (with detailed comparison attachment)3114. Feasibility, evidence sufficiency, and technical decomposition3125. Scoring and rating logic (with explicit rubric)3136. Stage-gate matrix (conditions, owners, timelines, non-fulfillment consequences)3147. Material gaps and data readiness3158. Counterevidence and weakening conditions316317Every claim must cite its source type and identifier, e.g., `[Patent: CN1234567B]`,318`[Paper: DOI or title]`, `[Web: source name]`, `[Proposal: section/page]`.319320## Formal Output Rules321322For formal review audiences such as large enterprises, public-sector bodies,323and research institutions:3243251. Remove empty opening filler such as trend-setting or era-setting lead-ins.3262. Remove promotional language and replace it with concrete review judgments.3273. Replace vague attribution with named evidence and explicit citations.3284. Rewrite formulaic paired constructions into direct standalone judgments.3295. Do not leak control-plane jargon such as `proposal-only`, `prior-art`,330 `assurance`, `gate`, or `owner` into the main report unless the term is331 necessary and clearly explained.332333Headings and body text should use plain formal wording suitable for a written334review or committee memo.335336When novelty is a live decision question, promote the method and point-by-point337comparison into `novelty-note.md`. A suitable file title is338`Public Novelty and Prior-Art Review Note` or `Novelty and Benchmarking Appendix`.339340## Completion Gates341342### All Modes343344- [ ] Concrete project object identified (not just a topic)345- [ ] Proposal-stated claims extracted before external search346- [ ] Four-layer evidence separation maintained throughout347- [ ] Tool tier and coverage limitations explicitly stated348- [ ] Material completeness assessment present349- [ ] If internal archives are unavailable, the duplicate-project boundary is marked as unverified350351### Screen Mode Additional352353- [ ] Go/no-go recommendation present with conditions354- [ ] Novelty boundary summary present355- [ ] Key gaps and next steps identified356357### Review Mode Additional358359- [ ] At least 2 patent searches and 1 paper search executed per innovation point360- [ ] Novelty comparison table present361- [ ] Every major claim has a traceable source citation362363### Innovation Mode Additional364365- [ ] Each innovation point assessed individually366- [ ] Differentiation boundary table present367- [ ] Residual differentiators explicitly stated per point368369### Assurance Mode Additional370371- [ ] Scoring rubric explicit and traceable372- [ ] Stage-gate matrix with conditions, owners, and consequences373- [ ] Counterevidence pass completed374- [ ] Technical decomposition covers main chain, dependencies, and failure modes375376## Guardrails377378- The proposal is an input object, not a source of truth. Do not let proposal379 rhetoric substitute for external evidence.380- Do not imply that a public-only retrieval pass proves exhaustive coverage or internal381 duplicate-project exclusion.382- Do not let raw search hit counts or internal ledger jargon become the main383 management-facing argument.384- Treat proposal promises, roadmap claims, and commercialization statements as385 signals until corroborated.386- Keep scoring as a support layer, not the main selling point. Prefer direct387 judgments (proceed / proceed with conditions / hold) over dense weighted scoring tables388 in the main body.389- When novelty is downgraded, do not stop at prose caution — build a structured390 differentiation-boundary table.391- Do not force a strong go/no-go when the evidence base is thin — prefer392 conditional recommendations with explicit next-evidence-gathering steps.393- Separate the novelty methodology (scope, queries, limitations) from the394 novelty conclusion.395- Do not mix proposal-stated claims with externally verified facts without396 clear labeling.397- For formal reports, do not leak control-plane jargon into the398 management-facing narrative.399- Treat internal duplicate-project exclusion as data-dependent — do not promise400 it when internal project history is unavailable.401402## Load These Files Only As Needed403404- [references/method-benchmark.md](references/method-benchmark.md)405- [references/domain-playbooks.md](references/domain-playbooks.md)406- [references/workflow.md](references/workflow.md)407- [references/source-routing.md](references/source-routing.md)408- [references/deliverables.md](references/deliverables.md)409- [references/evidence-schema.md](references/evidence-schema.md)410- [references/quality-gates.md](references/quality-gates.md)411- [templates/request-template.md](templates/request-template.md)412- [templates/workplan-template.md](templates/workplan-template.md)413- [templates/report-outline.md](templates/report-outline.md)