name: fireflies-transcript-analysis
description: Complete workflow for downloading Fireflies interview transcripts and analyzing them according to specified interview script. Use when user requests downloading transcripts, analyzing interviews, creating structured interview results, or working with Fireflies MCP transcripts.

Fireflies Transcript Analysis Workflow

Complete automation for downloading Fireflies interview transcripts and analyzing them according to a specified interview script.

Input Parameters

The workflow accepts two key parameters:

1. Meetings to download - Specify which meetings to process:
2. By transcript ID(s)
3. By date range
4. By participant name(s)
5. By meeting title keywords

6. All non-silent transcripts (default if not specified)

7. Interview script - Specify which script to use for analysis:

8. Path to script file (e.g., Internal-Wiki/02-product/research/ajtbd/ajtbd-interview-script.md)
9. Default: AJTBD script if not specified

Quick Start

When user requests transcript download and analysis:

1. Identify meetings - Determine which meetings to download based on user input
2. Download transcripts using fireflies-downloader subagent with meeting specifications
3. Analyze transcripts using transcript-analyzer subagent with script path
4. Save results in structured markdown format

Workflow Steps

Step 1: Identify Meetings to Download

Parse user input to determine which meetings to process:

Input formats:
- "Download transcripts with IDs: ABC123, DEF456"
- "Download all meetings with participant 'Vlad' from last week"
- "Download meetings from Jan 29, 2026"
- "Download all non-silent transcripts" (default)

Output: List of meeting identifiers (IDs, date ranges, participant names, etc.)

Step 2: Download Transcripts

Delegate to fireflies-downloader subagent with meeting specifications:

Use the fireflies-downloader subagent to download transcripts matching: [meeting specifications]

What the downloader does:
- Retrieves list of available transcripts from Fireflies MCP
- Filters transcripts based on provided criteria (IDs, dates, participants, etc.)
- Downloads specified transcripts in raw JSON format
- Saves with human-readable filenames: YYYY-MM-DD_respondent-name_interview.json
- Handles errors and provides feedback

Output: Raw JSON transcript files in Internal-Wiki/02-product/research/interviews/

Step 3: Identify Script for Analysis

Determine which interview script to use:

Input formats:
- "Analyze according to [script-path]"
- "Use AJTBD script" (default: Internal-Wiki/02-product/research/ajtbd/ajtbd-interview-script.md)
- "Analyze with custom script at [path]"

Output: Path to interview script file

Step 4: Analyze Transcripts

For each downloaded transcript, delegate to transcript-analyzer subagent with script path:

Use the transcript-analyzer subagent to analyze [transcript-file].json according to the script at [script-path].

What the analyzer does:
- Reads transcript JSON and specified interview script
- Extracts information for all script sections
- Records verbatim quotes (exact words, no paraphrasing)
- Creates structured markdown results following script structure
- Notes missing sections and interview quality

Output: Structured markdown files: YYYY-MM-DD_respondent-name_interview-results.md

Step 3: Verify Results

Check that:
- [ ] All transcripts downloaded successfully
- [ ] All transcripts analyzed
- [ ] Results files created with correct format
- [ ] Verbatim quotes included
- [ ] Missing sections noted

Rules and Protocols

This workflow uses:

1. Rule: fireflies-mcp-transcripts.mdc
2. MCP configuration
3. Download methods (MCP tool vs GraphQL API)
4. File naming conventions

5. Error handling

6. Protocol: ajtbd-transcript-analysis-protocol.mdc

7. Analysis workflow (Given-When-Then)
8. Script section mapping
9. Verbatim quote requirements

10. Result structure template

11. Subagents:

12. fireflies-downloader - specialized downloading
13. transcript-analyzer - specialized analysis

Example Usage

Download and Analyze Specific Meetings

User: Download transcripts with IDs ABC123 and DEF456, analyze them with AJTBD script

Workflow:
1. Identify meetings: IDs ABC123, DEF456
2. Use fireflies-downloader to download these specific transcripts
3. Use transcript-analyzer with AJTBD script path for each
4. Report completion with file locations

Download by Criteria and Analyze with Custom Script

User: Download all meetings with participant 'Vlad' from last week, analyze with custom script at Internal-Wiki/02-product/research/custom-script.md

Workflow:
1. Identify meetings: participant='Vlad', date range=last week
2. Use fireflies-downloader to filter and download matching transcripts
3. For each transcript, use transcript-analyzer with custom script path
4. Report summary: X transcripts downloaded, Y analyzed, Z results created

Analyze Existing Transcripts with Different Scripts

User: Analyze 2026-01-29_vlad_interview.json and 2026-01-29_yan_interview.json using the script at Internal-Wiki/02-product/research/ajtbd/ajtbd-interview-script.md

Workflow:
1. Skip download (files already exist)
2. For each transcript, use transcript-analyzer with specified script
3. Create results files for each

Multiple Scripts for Different Transcripts

User: Analyze transcript A with AJTBD script, transcript B with custom script at path/to/script.md

Workflow:
1. Download/identify transcripts A and B
2. Analyze transcript A with AJTBD script
3. Analyze transcript B with custom script
4. Report both results

File Locations

Transcripts (raw JSON):
- Internal-Wiki/02-product/research/interviews/YYYY-MM-DD_respondent-name_interview.json

Results (structured markdown):
- Internal-Wiki/02-product/research/interviews/YYYY-MM-DD_respondent-name_interview-results.md

Default Interview Script (AJTBD):
- Internal-Wiki/02-product/research/ajtbd/ajtbd-interview-script.md

Custom Scripts:
- Can be specified by user at any path
- Should follow similar structure with sections/questions

Error Handling

MCP unavailable:
- Fall back to GraphQL API via curl
- Inform user about workaround

Missing transcript data:
- Note in analysis which sections are missing
- Assess interview quality
- Suggest if additional interview needed

File save errors:
- Use /tmp/ as intermediate storage
- Retry copy operation
- Report permission issues

Key Principles

1. Delegate to specialists - Use subagents for their specific tasks
2. Follow protocols - Adhere to rules and protocols
3. Preserve verbatim - Never paraphrase respondent words
4. Be thorough - Check all script sections
5. Clear communication - Report progress and issues

Checklist

Before completing workflow:

• [ ] Transcripts downloaded (if requested)
• [ ] Transcripts analyzed (if requested)
• [ ] Results files created with correct format
• [ ] Verbatim quotes included
• [ ] Missing sections noted
• [ ] Interview quality assessed
• [ ] Files saved in correct locations
• [ ] User informed of completion

Next Steps After Analysis

After creating results:
- Results are ready for review
- Can be used for product decisions
- Missing sections may require follow-up interviews
- Verbatim quotes can inform messaging and positioning