메뉴
Create or analyze Developer Experience surveys. Creation mode outputs a 9-question HaTS-aligned survey; Analysis mode outputs sentiment metrics with theme clustering.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Help users document system architecture with Mermaid diagrams; produces overview, tier deep-dives, data flows, and a component summary table.
Invoke when planning a user interview; produces a best-practice template (Open, Usability, or Mixture) with Mom Test-aligned questions.
Help users prepare and deliver decision briefings using the SOCRR framework. Use when someone needs a decision from a manager, executive, sponsor, or stakeholder group.
Help users define user needs with personas, a use case table (Outcome / How Today / Future / North Star), and open questions.
Help users craft persuasive narratives; produces a Story Map with headline, opening, narrative arc, and delivery practice guide.
Write a complete Product Requirements Document (PRD) for a feature, product area, or initiative. Trigger this skill whenever someone asks to write a PRD, requirements doc, spec, or product brief — or says things like "I need to document this feature," "help me write up what we're building before sprint planning," or "my VP wants a spec for this project." This is the default PRD format covering problem statement, goals, user stories, success metrics, technical considerations, milestones, and risks — designed to align engineering, design, TPM, and leadership. Do NOT trigger for: one-pagers for exec approval (use create-prd-one-pager), specs optimized for AI coding agents (use create-prd-claude-code), deep engineering specs with API contracts (use create-prd-engineering), or build-vs-buy decisions (use decision-brief).
| name | devex-survey |
| description | Create or analyze Developer Experience surveys. Creation mode outputs a 9-question HaTS-aligned survey; Analysis mode outputs sentiment metrics with theme clustering. |
This skill defines how AI assists with Developer Experience surveys in two modes:
Both modes enforce strict constraints on question design, statistical reporting, and guardrails against misuse.
See references/runbook.md for the full operational playbook covering distribution, cadence, and communication strategy.
This skill exists because developer experience measurement requires survey science rigor (HaTS alignment, Likert scales, longitudinal consistency). Ad-hoc surveys produce unreliable data that can't be compared across quarters.
| Artifact | Format | Handed to |
|---|---|---|
| Survey (Creation mode) | 9-question survey with scales and rationale | Survey platform / PM |
| Analysis report (Analysis mode) | Markdown with sentiment metrics, themes, leverage opportunities | PM / leadership |
When the user invokes this skill, determine which mode applies:
Generate a quarterly Developer Experience survey that:
Incorporate principles from:
These constraints are non-negotiable. Every generated survey must comply:
| Constraint | Requirement |
|---|---|
| Max questions | 9 total (5 core + 1 pain point + 1 consolidated why + 2 floaters) |
| Scale | 5-point Likert (see below for allowed scales) |
| Phrasing | Non-biased, conversational (see below) |
| Identity | Non-anonymous |
| Completion time | 3-5 minutes |
| Layout | Single page |
| Floater questions | 1-2 allowed |
| Core questions | Must remain stable quarter-over-quarter |
Satisfaction scale (default for core questions):
Ease/difficulty scale (allowed for floater questions where it fits better):
Floater questions may use either scale depending on what is being measured. Core questions always use the satisfaction scale.
Satisfaction % = (Very Satisfied + Satisfied) / Total Responses
Dissatisfaction % = (Very Dissatisfied + Dissatisfied) / Total Responses
Neutrals remain in the denominator. Do NOT generate vanity CSAT scores.
Every survey must include these 5 sentiment questions. They must remain identical across quarters to preserve longitudinal integrity.
Each question uses the phrasing pattern:
"How satisfied or dissatisfied are you with..."
Write questions the way a developer would talk — use "your code", "merging your code", not abstract process language.
Include exactly one pain point selection question:
"What are your top 3 pain points?"
After the core questions and pain point selection, include a single open-ended question:
"Why did you respond the way you did to the above questions?"
This replaces per-question follow-ups. One consolidated "why" reduces survey fatigue while still capturing the reasoning behind sentiment responses.
Floater questions are contextual and may rotate each quarter. Use them to assess:
Floater questions do NOT replace core questions. Core questions are always present and unchanged.
Floaters may use the satisfaction scale or the ease/difficulty scale depending on what is being measured. For example, a question about stewardship approval friction fits the ease/difficulty scale better than satisfaction.
Do NOT ask respondents for:
These dimensions are available via metadata and will be joined during analysis. Asking for them wastes question slots and increases survey fatigue.
The full survey follows this order:
When generating a survey, return:
Analyze survey results to produce:
Join survey responses with organizational metadata to segment by:
Identify:
Flag respondents who:
Always calculate and report:
| Metric | Required |
|---|---|
| Total invited | Yes |
| Total responded | Yes |
| Response rate | Yes |
| Confidence interval | Yes |
| Satisfaction % per question | Yes |
| Dissatisfaction % per question | Yes |
Target CI: 1-3%. Never publish sentiment without confidence interval.
Return a structured report with these sections:
These rules apply to both modes and must never be violated:
The goal is trust, not vanity metrics.
/analyzing-user-feedback — General user feedback analysis frameworks/designing-surveys — Broader survey design principles/storytelling-for-impact — Craft compelling narratives from survey data for shareouts/giving-presentations — Present survey findings at all-hands and leadership reviews