crisp-mcp

name	crisp-mcp
description	Use the CRISP‑T MCP server to run a full qualitative analysis workflow linking textual narratives and numeric data.

Summary

Goal: Use the CRISP‑T MCP server to run a full, documented qualitative analysis workflow—linking textual narratives and numeric data—using computational triangulation and grounded theory–inspired coding.

This skill assumes:

• The CRISP‑T MCP server is available as an MCP server called crisp.
• If you do not find the crisp MCP server, inform the user and use crisp-cli skill instead.
• The server exposes several tools for corpus management, text coding, numeric analysis, triangulation, and export
• The model can discover tools at runtime and adapt to the actual tool list

Use this skill when the user wants to:

• Explore and code qualitative data (e.g., interviews, narratives, open‑ended survey responses)
• Triangulate qualitative codes with quantitative variables
• Produce transparent, reproducible outputs (memos, codebooks, matrices, summaries)

---

Intended users and scenarios

• Qualitative / mixed‑methods researchers working with text + numeric data

Typical scenarios:

• Coding COVID‑era patient narratives and linking them to open‑source epidemiological data
• Triangulating interview themes with survey scores or EHR‑derived indicators
• Building theory from mixed text–numeric corpora using CRISP‑T stages

---

MCP server and tools

Server

• Server name (example): crisp-t
• Protocol: Model Context Protocol (MCP) via stdio (e.g., crisp-mcp)

Important: Use ./workspace as the working directory for all artifacts, if not explicitly specified. Create it if it does not exist. Always discover the available tools from the CRISP‑T MCP server first, then map them to the workflow below. Tool names in this file are illustrative; prefer the actual tool names and schemas exposed by the server.

Recommended steps

• Import data into CRISP-T corpus and dataframe.
• Perform linking between text and numeric data using various methods (id based, keyword based, time based, embedding based).
• Explore text data using various methods (e.g., topic modeling, keyword extraction, sentiment analysis, visualizations).
• Explore numeric data using various methods (e.g., summary statistics, classification, clustering, regression, association, visualizations, TDA, etc.).
• Perform cross modal analysis using linked text and numeric data (e.g., text features as predictors for numeric outcomes, numeric features as predictors for text outcomes, etc.).
• Add manual connections between text documents and numeric rows if needed to support theory driven analysis.
• Derive insights from the analysis and document them.

Expected tool capabilities (illustrative)

When you discover tools from the crisp-t MCP server, look for tools that correspond to the following capabilities:

1. Project and configuration

   • List / load projects (e.g., list_projects, load_project)
   • List / configure data sources (text + numeric)

2. Textual data operations

   • List text sources (e.g., documents, transcripts, narrative files)
   • Fetch text segments (e.g., by document, case, or segment ID)
   • Apply or update codes to segments
   • List codes / categories / codebook

3. Numeric / structured data operations

   • List numeric datasets (e.g., CSVs, tables)
   • Fetch rows / variables for a given case or group
   • Run basic summaries or models (e.g., descriptive stats, correlations)

4. Triangulation / integration

   • Link coded text segments to numeric data by case ID or key
   • Generate triangulation matrices (codes × variables, cases × themes, etc.)
   • Compute simple indicators (e.g., frequency, co‑occurrence, association)

5. Memoing and export

   • Create / update analytic memos
   • Export codebooks, matrices, and summaries (e.g., CSV, JSON, Markdown)
   • Save session context or analysis snapshots

When using this skill, you should:

• Inspect tool descriptions and schemas
• Prefer higher‑level CRISP‑T tools (e.g., “triangulate”, “inspect”, “sense‑make”) over low‑level primitives when available
• Avoid guessing tool parameters—use the schema and ask the user when needed

---

Inputs

The user may provide some or all of the following. If missing, ask concise clarification questions.

• Research question / focus
  • research_question: short description of what they want to understand or explain
• Project / dataset selection
  • project_name or project_id
  • text_source_pattern (e.g., “patient_narratives”, “interviews_*”)
  • numeric_dataset_name (e.g., “covid_metrics.csv”, “survey_scores”)
• Methodological preferences
  • coding_style: open coding, focused coding, or using an existing codebook
  • triangulation_focus: e.g., “compare themes by age group”, “link narratives to hospitalization status”
• Output preferences
  • output_format: Markdown, CSV, JSON, or mixed
  • audience: “methods‑savvy colleague”, “student”, “grant reviewer”, etc.

If the user is vague, default to:

• Open coding → focused coding
• Simple triangulation (themes × one or two key numeric variables)
• Markdown summaries plus CSV exports where supported

---

Outputs

This skill should aim to produce:

1. Analytic memo

   • Research question and context
   • Data sources used (text + numeric)
   • Coding approach and key categories
   • Triangulation findings (convergences, divergences, surprises)
   • Limitations and next steps

2. Structured artifacts (when tools support them)

   • Codebook (codes, definitions, examples)
   • Triangulation matrix (e.g., codes × variables)
   • Exported files (CSV/JSON/Markdown) referenced in the memo

3. Reproducible trace

   • Clear description of which MCP tools were called
   • Parameters used (e.g., filters, variables, thresholds)
   • Any important analytic decisions made during the session

---

High‑level workflow

You should follow this workflow, adapting to the actual CRISP‑T tools you discover.

1. Clarify research focus and context

1. Summarize the user’s goal in 2–3 sentences.
2. If unclear, ask one or two focused questions:
   • Example: “Which is more important right now: exploring themes, or linking themes to numeric outcomes?”
3. Record the clarified research question in the analytic memo.

2. Discover CRISP‑T MCP tools and project context

1. List available tools from the crisp-t MCP server.
2. Identify tools that correspond to:
   • Project management
   • Text operations
   • Numeric operations
   • Triangulation
   • Memoing / export
3. If available, list projects and either:
   • Load the project specified by the user, or
   • Present a short list and ask the user to choose.

Document in the memo:

• Which project was loaded
• Which tools will be used and for what purpose

3. Inspect data sources (text + numeric)

1. Use project / data tools to:
   • List text sources (e.g., narrative collections, interview sets)
   • List numeric datasets (e.g., survey tables, EHR‑derived tables)
2. For each candidate source:
   • Fetch a small sample (e.g., a few narratives, a few rows of numeric data)
   • Briefly describe:
     • What the data represent
     • How they relate to the research question
3. Confirm with the user which sources to prioritize.

Update the memo with:

• Selected text sources and why
• Selected numeric datasets and why

4. Initial coding of qualitative data (open / exploratory)

1. Select a manageable subset of text (e.g., 10–20 narratives or segments).
2. Apply open coding using the CRISP‑T coding tools:
   • Generate short, grounded codes that stay close to the data
   • Avoid premature theory; focus on what participants are saying/doing/experiencing
3. Use MCP tools to:
   • Save codes to the project
   • Attach codes to segments
   • Retrieve and refine the emerging code list

In the memo:

• Describe the coding approach (e.g., open coding, line‑by‑line, incident‑focused)
• Provide examples: quote → code → brief interpretation

5. Focused coding and category development

1. Review the emerging code list using CRISP‑T tools (e.g., list codes, code frequencies).
2. Group related codes into higher‑level categories or themes:
   • Merge overlapping codes
   • Distinguish closely related but conceptually different codes
3. If the server supports it, create or update:
   • A codebook (code name, definition, inclusion/exclusion criteria, example quotes)
   • Category structures (e.g., parent/child relationships)

In the memo:

• Describe key categories and how they were derived
• Note any tensions, ambiguities, or alternative interpretations

6. Computational triangulation with numeric data

Using CRISP‑T tools that link text and numeric data:

1. Identify the linkage key (e.g., participant ID, case ID, encounter ID).
2. For each case or group:
   • Retrieve coded qualitative data (codes, categories, or themes)
   • Retrieve numeric variables (e.g., age, scores, outcomes, counts)
3. Construct triangulation views, such as:
   • Code frequencies by group (e.g., theme prevalence by age band or outcome)
   • Associations between codes and numeric indicators (e.g., high vs. low scores)
   • Patterns of convergence (where qualitative and quantitative tell a similar story)
   • Patterns of divergence (where they conflict or complicate each other)

When tools support it, generate:

• Triangulation matrices (e.g., codes × variables, cases × themes)
• Summary statistics (e.g., counts, proportions, simple associations)

In the memo:

• Describe how triangulation was done (data sources, linkage, variables)
• Highlight convergences, divergences, and surprises
• Emphasize that triangulation is about enriching and challenging interpretations, not “proving” them

7. Synthesis and theory‑building

1. Integrate insights from:
   • Qualitative categories and themes
   • Numeric patterns and contrasts
   • Triangulation matrices and summaries
2. Propose tentative theoretical statements or propositions:
   • “In this context, participants who X tend to Y, especially when Z…”
3. Explicitly note:
   • Alternative explanations
   • Data limitations
   • What additional data or analysis would strengthen the claims

Update the memo with:

• A coherent narrative that links data, codes, categories, and numeric patterns
• Clear statements about what the data support and what remains uncertain

8. Export and documentation

1. Use CRISP‑T export tools to save:
   • Codebook(s)
   • Triangulation matrices
   • Any generated summaries or tables
2. If supported, export:
   • The analytic memo as Markdown or another requested format
   • A machine‑readable representation of the analysis (e.g., JSON snapshot)

Provide the user with:

• A human‑readable summary (memo)
• A list of exported artifacts and where they are stored (paths, IDs, or URLs)

---

Model behavior and style guidelines

When using this skill, you should:

• Be methodologically explicit
  • Name what you are doing (open coding, focused coding, triangulation, memoing)
  • Briefly justify key analytic decisions
• Stay grounded in the data
  • Use short quotes and concrete examples
  • Avoid over‑generalizing beyond what the data support
• Respect the user’s context
  • If they are teaching, highlight steps and rationale for students
  • If they are writing a paper or grant, emphasize clarity and traceability
• Limit questions
  • Ask at most one or two focused questions at a time
  • Default to reasonable assumptions when the user is not specific

---

Limitations and cautions

• Triangulation results should be treated as aids to interpretation, not definitive proof.
• Always encourage the user to:
  • Review codes and categories
  • Challenge interpretations
  • Consider ethical and contextual factors not visible in the data

---