Agent Skills
› mohitagw15856/pm-claude-skills
› user-interview-synthesis
user-interview-synthesis
GitHub将用户访谈原始记录转化为结构化研究报告，提炼主题、痛点及可执行洞察。适用于分析访谈笔记、综合定性研究或从数据中提取产品建议，输出含引用、启示及下一步行动。
触发场景
分析访谈笔记
综合定性研究
从访谈中识别主题
将原始访谈数据转化为产品洞察
安装
npx skills add mohitagw15856/pm-claude-skills --skill user-interview-synthesis -g -y
SKILL.md
Frontmatter
{
"name": "user-interview-synthesis",
"description": "Synthesises user interview transcripts into structured research findings. Use when asked to analyse interview notes, synthesise qualitative research, identify themes from interviews, or turn raw interview data into actionable product insights. Produces a themed synthesis with supporting quotes per theme, 'so what' implications, and recommended next steps. For mixed sources beyond interviews (surveys, tickets, feedback) use user-research-synthesis instead."
}
User Interview Synthesis Skill
Transform raw interview transcripts into a structured synthesis document that surfaces themes, pain points, and actionable insights.
Required Inputs
Ask the user for these if not provided:
- Interview transcripts or notes (even rough notes work)
- Number of participants and their profiles (role, company size, context)
- Research questions (what was the study trying to answer?)
- Date range of research (for context)
Process
- Read all provided transcripts fully before drawing conclusions
- Identify recurring themes (minimum 3 mentions to qualify as a theme)
- Categorize findings into: Pain Points, Workflow Insights, Feature Requests, Delight Moments
- Select 2-3 verbatim quotes per theme that best represent the pattern
- Draft "So What" implications for each theme — what does this mean for the product?
- Validate — Confirm every theme has quotes from at least 3 participants. Flag any insight resting on fewer as low-confidence.
Output Structure
Research Synthesis: [Study Name]
Participants: [n] Date Range: [dates] Research Questions: [list]
Theme 1: [Theme Name]
- Summary (2-3 sentences)
- Supporting quotes (from at least 3 participants)
- Implication for product
[Repeat for each theme]
Low-Confidence Signals (1-2 participants only)
[Findings worth tracking but not acting on yet — note what further research would confirm or deny]
Recommended Next Steps
[Specific, actionable recommendations based on findings]
Deeper Materials
This skill ships with support files — use them when they are available:
references/coding-transcripts.md— Coding Interview Transcripts Without Losing the Signal. Apply it while producing the output; it carries the calibration and judgment calls the method summary above compresses.templates/per-session-capture.md— a fill-in version of the deliverable with the quality gates inline. Offer it when the user wants to work the document themselves rather than have it generated.
Quality Checks
- Every theme is supported by quotes from at least 3 participants
- Implications connect to specific product decisions, not just observations
- Researcher bias check: no leading language, findings don't all support one hypothesis
- Single-source signals are flagged separately, not mixed into main themes
- Research questions from the study brief are each addressed (even if the answer is "inconclusive")
Anti-Patterns
- Do not mix single-source signals into main themes — insights cited by only one participant must be flagged separately
- Do not write implications that are observations restated rather than product decisions enabled
- Do not include themes that only support the project hypothesis — contradictory findings must be surfaced, not omitted
- Do not present findings without quotes — every theme requires verbatim evidence from at least 3 participants
- Do not leave research questions unanswered — each question from the study brief must be explicitly addressed, even if the answer is inconclusive
版本历史
- a38bc30 当前 2026-07-05 11:47
