filedata
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-11-11
- Stars
- ★ 16
- Forks
- 25
03 SKILL.md 正文
Purpose
Use Dataverse MCP to perform duplicate checks and save the use-case and assessment records.
When to use
Use only after the user explicitly chooses Submit and the Confirmation and Final QA skill returns readyToSubmit = true.
Tool
Use Dataverse MCP only.
Target Dataverse solution and tables
Target solution:
- AI Use Case Solution
Target tables:
- Use-case table display name: AI_UseCase
- Use-case table logical name: cadpbm_AI_UseCase
- Use-case primary ID: cadpbm_AI_UseCaseId
- Assessment table display name: AI Assessment
- Assessment table logical name: cadpbm_AIAssessment
- Assessment primary ID: cadpbm_AIAssessmentId
- Agent App Catalog URL environment variable schema name: cadpbm_CodeAppURL
These names are internal implementation details. Do not include them in normal user-facing responses.
Inputs
- Confirmed use-case data.
- Confirmed assessment data.
- Submit confirmation.
Required behavior
- Retrieve target table metadata/schema silently.
- Retrieve category and subcategory choice values silently.
- Check for duplicate Title or Summary in the last 60 days.
- If duplicate found, return a user-safe match and request Proceed, Update existing, or Cancel.
- If no duplicate, proceed silently.
- Create or update the use-case record.
- Retrieve the internal record reference silently.
- Create or update the assessment record.
- Associate the assessment with the use-case record.
- Retrieve the environment variable value for cadpbm_CodeAppURL.
- Return human-readable IDs and the Agent App Catalog URL.
ID formats
- Use-case human-readable ID: UC-{ddmmyyhhmm}
- Assessment human-readable ID: ASMNT-{ddmmyyhhmm}
Fields to save
Use-case record:
- Customer Name
- Idea Title
- Category
- Subcategory
- Summary
- Problem / Need
- Proposed Solution
- Expected Benefits
- Target Audience
- Personas Impacted
- Potential Risks
- Sample Prompts, Agent Questions, and Example Answers
Assessment record:
- Every impact factor score and rationale
- Every complexity factor score and rationale
- Automation and AI Usage
- Primary Work Type
- Expertise and Tuning
- Risk and Control Profile
- Reach Score
- Agentic Fit
- Impact Average
- Complexity Average
- CAD Intake Quality
- Weak spots or validation notes
Rules
- Do not save unless Submit is confirmed.
- Do not expose table names, schemas, columns, choice internals, or GUIDs to the user.
- Do not ask the user to provide or paste a Dataverse GUID.
- For existing-record lookup, ask for the human-readable use-case ID, assessment ID, title, or customer name.
- Map category and subcategory through Dataverse choice values.
- Retry once on save failure.
- Never fail silently.
- When reporting progress to the user, use short icon-prefixed updates such as "? Checking for similar recent records", "?️ Saving the use case", and "✅ Use case saved".
- Do not expose raw duplicate-check logic, schema details, choice option internals, internal record references, or GUIDs in progress updates.
- If choice mapping is ambiguous, explain it as a user-safe decision point rather than exposing raw option values.
- Retrieve the Agent App Catalog link from cadpbm_CodeAppURL after successful save.
- Do not hardcode the app catalog URL.
- Do not save the app catalog URL or generated link back into Dataverse.
- If cadpbm_CodeAppURL is missing or empty, return a plain-language warning that the App Catalog link is not configured for this environment.
Output contract
{
"duplicateFound": false,
"duplicateMatchUserSafe": "",
"requiresDuplicateDecision": false,
"useCaseId": "",
"assessmentId": "",
"agentAppCatalogUrl": "",
"saved": false,
"plainLanguageError": ""
}
04 同仓库更多 Skills
dock-assignment-pass-pdf
Render the completed truck loading-dock assignment as a driver-facing one-page PDF.
loading-dock-selector
Reason over weather, one current dock snapshot, optional ETA, and verified Drive Priority Pass to select and explain one loading dock.
newsletter-style-email-composer
Use when the user asks to email a report, findings, or metrics to someone as a polished, newsletter-style message. Builds an Outlook-safe HTML email from figures that already exist (headline, KPI tiles, up to two bar charts drawn with HTML table cells so they render in classic Outlook, Gmail and dark mode, short findings, data notes and the non-audit disclaimer) using the shared brand tokens, attaches the PDF report, and creates a draft for the user to review. Do NOT use for a short plain-text status update, or for building the underlying analysis, charts, or report itself -- this skill only composes the email.
polished-report-builder
Use when the user asks to turn analysis results, metrics, or findings into a polished, presentation-ready report, one-pager, or executive summary -- especially when existing analysis (e.g. from the spreadsheet analysis skill) needs to become a document someone will actually read. Builds a branded PDF with reportlab from a short content file (headline, KPI tiles, one chart per point, data notes and the non-audit disclaimer) using the shared brand tokens, and runs a QA pass before delivery to catch walls of text, chart clutter, low-resolution charts, missing data notes or poor contrast. Do NOT use for a quick inline chat answer, a single ad-hoc chart with no report structure around it, a raw data export with no narrative, or to compute the figures themselves.
spreadsheet-data-analysis-reporting
Use for any request that needs figures from the sales spreadsheets (Excel/CSV), even a single file or a single period - e.g. 'analyse Q4', 'how did sales do last month', 'compare regions', YoY/MoM/QoQ/YTD growth, top products, share of revenue, trends, outliers, currency-converted totals, or building/updating a consolidated master file. Finds and loads the relevant files, cleans and normalises them, computes metrics with standard definitions and edge-case handling, flags outliers and data issues, explains every transformation, and produces on-brand charts. Use this before the report or email skills whenever numbers must be computed. Do NOT use for questions that need no data (e.g. 'what does YoY mean') or for re-formatting results already computed in this conversation.
generating-podcast-script
Use this skill whenever the user asks to write, generate, or create a podcast script or podcast episode — from a topic, or from source material such as a news digest, newsletter, email review, or set of articles — and optionally convert it to audio with Azure Text-to-Speech. Handles the initial request and every follow-up refinement (source, topic, length, cast, narration) in the same task.