Agent-Skills.md

Agent Skills: Analyze Transcript Command

Analyze meeting transcripts to extract requirements, decisions, and action items. Specialized extraction for conversational content.

UncategorizedID: melodic-software/claude-code-plugins/analyze-transcript

Repository

melodic-software/claude-code-plugins
melodic-softwareLicense: MIT
417

Install this agent skill to your local

pnpm dlx add-skill https://github.com/melodic-software/claude-code-plugins/tree/HEAD/plugins/requirements-elicitation/skills/analyze-transcript

Skill Files

Browse the full folder contents for analyze-transcript.

Download Skill

Loading file tree…

plugins/requirements-elicitation/skills/analyze-transcript/SKILL.md

Skill Metadata

Name
analyze-transcript
Description
Analyze meeting transcripts to extract requirements, decisions, and action items. Specialized extraction for conversational content.

Analyze Transcript Command

Extract requirements and decisions from meeting transcripts and conversational content.

Usage

/requirements-elicitation:analyze-transcript ./meetings/kickoff-notes.md
/requirements-elicitation:analyze-transcript ./transcripts/stakeholder-call.txt --domain "checkout"
/requirements-elicitation:analyze-transcript ./notes/*.md --speakers "PM:ProductManager,TL:TechLead,UX:Designer"

Arguments

| Argument | Required | Description | |----------|----------|-------------| | transcript-path | Yes | Path to transcript file or glob pattern | | --domain | No | Domain name for organizing output | | --speakers | No | Speaker role mapping (format: abbreviation:Role) |

What It Extracts

Requirements

  • Explicitly stated requirements
  • Implied requirements from discussions
  • Feature requests and suggestions

Decisions

  • Agreed-upon approaches
  • Technology choices
  • Scope decisions

Action Items

  • Commitments made
  • Follow-up tasks
  • Assignments

Concerns

  • Risks identified
  • Issues raised
  • Open questions

Extraction Patterns

Decision Markers

"We decided that..."
"The conclusion is..."
"Going with [option]..."
"Agreed: ..."
"Decision: ..."

Action Item Markers

"Action: [name] will..."
"TODO: ..."
"[Name] to follow up on..."
"We need to..."
"Next step: ..."

Requirement Discussions

"The system should..."
"Users need to be able to..."
"It must handle..."
"Requirements: ..."
"Feature request: ..."

Concern Markers

"I'm concerned about..."
"Risk: ..."
"Issue: ..."
"What if...?"
"Problem: ..."

Workflow

Step 1: Load Transcript

Read the transcript file(s) using the Read tool.

Step 2: Identify Speakers

If --speakers provided, use the mapping. Otherwise, attempt to identify speakers from patterns like:

  • [Name]: or Name:
  • **Name**:
  • Speaker labels in transcript format

Step 3: Extract Content

Spawn document-miner agent with transcript-specific extraction patterns.

Extract:

  1. Requirements (explicit and implicit)
  2. Decisions
  3. Action items
  4. Concerns and risks

Step 4: Attribute and Weight

For each extraction:

  • Note the speaker (if identifiable)
  • Apply weighting based on speaker role
  • Record surrounding context

Step 5: Save and Report

Save to .requirements/{domain}/transcripts/ Report summary of findings.

Speaker Weighting

speaker_weights:
  product_owner: high    # Business requirements
  technical_lead: high   # Technical constraints
  stakeholder: high      # Domain requirements
  developer: medium      # Implementation insights
  designer: medium       # UX requirements
  unknown: low           # Needs validation

Examples

Basic Transcript Analysis

/requirements-elicitation:analyze-transcript ./meetings/sprint-planning.md

Output:

Analyzing: sprint-planning.md

Speakers identified:
- PM (ProductManager): 45 statements
- TL (TechLead): 32 statements
- DEV (Developer): 28 statements

Extraction Results:

REQUIREMENTS (8):
- REQ-TR-001: "Users shall receive email notifications for order status" [PM, High]
- REQ-TR-002: "API response time must be under 200ms" [TL, High]
...

DECISIONS (3):
- DEC-001: "Use PostgreSQL for primary database" [TL]
- DEC-002: "Target launch date: Q2 2025" [PM]
...

ACTION ITEMS (5):
- ACT-001: "TL to create database schema document"
- ACT-002: "PM to get stakeholder sign-off"
...

CONCERNS (2):
- CONCERN-001: "Third-party API reliability" [TL]
- CONCERN-002: "Training timeline for new features" [PM]

Saved to: .requirements/sprint-planning/transcripts/TR-sprint-planning.yaml

With Speaker Mapping

/requirements-elicitation:analyze-transcript ./calls/client-call.txt \
  --speakers "JD:ClientCEO,SM:SalesManager,PM:ProductManager" \
  --domain "enterprise-deal"

Output:

Analyzing: client-call.txt

Speaker Mapping Applied:
- JD → ClientCEO (weight: high)
- SM → SalesManager (weight: medium)
- PM → ProductManager (weight: high)

REQUIREMENTS from ClientCEO (5):
- REQ-TR-001: "Must integrate with SAP" [JD, High confidence]
- REQ-TR-002: "Support 50,000 concurrent users" [JD, High confidence]
...

REQUIREMENTS from ProductManager (3):
- REQ-TR-006: "Dashboard customization needed" [PM, High confidence]
...

Saved to: .requirements/enterprise-deal/transcripts/TR-client-call.yaml

Multiple Transcripts

/requirements-elicitation:analyze-transcript ./meetings/*.md --domain "q1-planning"

Output:

Found 4 transcripts matching pattern

Processing:
1. kickoff-meeting.md ........ 12 requirements, 4 decisions
2. technical-review.md ....... 8 requirements, 6 decisions
3. stakeholder-feedback.md ... 15 requirements, 2 decisions
4. sprint-planning.md ........ 5 requirements, 3 decisions

Cross-Transcript Analysis:
- Total requirements: 40
- Duplicates detected: 6 (consolidated to 34)
- Conflicting statements: 2 (flagged for review)

Summary saved to: .requirements/q1-planning/transcripts/summary.yaml

Output Format

Saved YAML Structure

transcript_analysis:
  file: "sprint-planning.md"
  analyzed_date: "2025-12-25T14:30:00Z"
  domain: "{domain}"

speakers:
  - abbreviation: PM
    role: ProductManager
    statement_count: 45
    weight: high
  - abbreviation: TL
    role: TechLead
    statement_count: 32
    weight: high

requirements:
  - id: REQ-TR-001
    text: "System shall send email notifications for order status changes"
    speaker: PM
    speaker_role: ProductManager
    context: "Discussion about customer communication"
    original_quote: "We need email notifications when orders change status"
    confidence: high
    type: functional

decisions:
  - id: DEC-001
    text: "Use PostgreSQL for primary database"
    speaker: TL
    context: "Database technology selection"
    rationale: "Team expertise and scalability needs"

action_items:
  - id: ACT-001
    text: "Create database schema document"
    assignee: TL
    due_date: null  # If not specified
    context: "Follow-up from database decision"

concerns:
  - id: CONCERN-001
    text: "Third-party API reliability during peak hours"
    raised_by: TL
    context: "Integration discussion"
    severity: medium

conflicts:
  - items: [REQ-TR-003, REQ-TR-015]
    description: "Conflicting performance targets"
    resolution_needed: true

Integration

Follow-Up Commands

# Check for gaps after transcript analysis
/requirements-elicitation:gaps

# Interview stakeholders about concerns raised
/requirements-elicitation:interview "Technical Lead" --context "API reliability concerns"

# Consolidate with other sources
/requirements-elicitation:discover "{domain}" --sources transcripts,documents

Best Practices

Transcript Preparation

For best results, transcripts should:

  • Clearly identify speakers
  • Use consistent formatting
  • Include timestamps (optional but helpful)
  • Separate topics with headers

Review Recommendations

After analysis:

  1. Review all flagged conflicts
  2. Validate low-confidence extractions
  3. Follow up on open concerns
  4. Assign action items if not already done