Survey Feedback Synthesizer skill

[Tim (Timziito)]

Turns open-text survey, NPS, or event feedback into themes, sentiment, and a prioritized action list, grounded only in what respondents actually wrote, honest about sample size and outliers.

[Copilot]

Pull request overview

Adds a new Agent Skill submission that synthesizes open-text survey/NPS/event feedback into themes, sentiment, and a prioritized action list, with guardrails against over-inference and over-weighting outliers.

Changes:

• Added survey-feedback-synthesizer skill instructions with explicit guardrails and tone guidance.
• Added a human-facing README describing the skill’s intent and constraints.
• Added submission metadata (platforms/tags/author/version/dates) for gallery ingestion.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
submissions/survey-feedback-synthesizer/SKILL.md	Defines the agent-facing procedure, guardrails, and tone for synthesizing feedback.
submissions/survey-feedback-synthesizer/README.md	Provides human-facing overview of what the skill does and doesn’t do.
submissions/survey-feedback-synthesizer/metadata.json	Registers the submission with required metadata (name/description/platforms/tags/author/version/dates).

[Tim (Timziito)]

@microsoft-github-policy-service agree

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

[Adriana Trujillo (adrianatruji)]

Thank you, Tim (@Timziito) ! Really like seeing customer feedback and survey data in a proposal. The thing I'd push on is differentiation: As written now, the synthesis function of the skill - themes, sentiment, action notes - is close to what specialized agents like Analyst are meant to do with attached files of this type, so I would think about extended capabilities to bring some unique added value.
Where it could genuinely earn its place is by leaning into what these platforms (Cowork, Copilot Studio, Scout) can do using their broad sandbox/code capabilities to turn the analysis into an actual deliverable the user can tailor (e.g. a dashboard-type file the user can brands with their own visual assets / company guidelines). That shifts the value from "analyze the feedback" (already owned) to "produce a unique, on-brand artifact off the back of that analysis," which is a real, reusable procedure. Just an example, but keen to review a version re-framed to account for that uniqueness! Thank you for all the ideas, appreciate it!