Customer Research (voice-of-customer synthesis)

The everyday workhorse: someone asks a question about customers and gets back not a plausible guess but a quote-rich, sourced brief drawn from real conversations. This is the skill that most visibly changes the agent's behavior after Evermuse is installed — so lean into the evidence.

Step 0 — Relevance & availability

Confirm it's a customer question and Evermuse is connected. If not connected, say so plainly and tell the user to authorize the MCP — don't answer a "what do customers think" question from memory and pass it off as grounded.

Scope the question

Pin down what's really being asked: a topic ("onboarding"), a decision it feeds ("should we build X"), a segment or time window. If a research project is clearly implied (e.g. "the audio-listeners study"), pass its project_id on your calls; otherwise stay at product scope.

Evermuse Grounding (required)

Follow ${CLAUDE_PLUGIN_ROOT}/skills/using-evermuse/SKILL.md.

Required — one parallel batch of searches. Verify the product (Rule 1), then fire in a single parallel batch:

• 3–4 evidence searches, each worded from a different angle (limit up to 50). Evidence comes back rich and varied — expect large, useful result sets.
• one guidance search and one context search (limit up to 50). These are usually sparse or empty; run them anyway and note when they're thin.

Read each response's digest — it reports how many more results exist. Use judgment on whether a query is worth pulling deeper (raise limit toward the 100 max and/or page with next_offset to avoid repeats), weighing payload size, remaining context, task complexity, and the value of the data. For deep pulls, consider spawning sub-agents — instruct them to return every citation with the same metadata the tools return (url, who_said_it, meeting_name, created_at) so you can still cite.

Optional — considered use. Once grounded, reach for the other tools only when they add value: a quote-angled search for verbatim voice, read_source for a single deep dive, inline citations (references/citations.md), and add_source to save the deliverable. Sources, citations, and saving are optional — not required.

Synthesize

Cluster the evidence into themes, and for each: demand strength (mentions across distinct accounts), sentiment split, who said it and when, and the sharpest verbatim quote. Represent dissenting voices — don't flatten disagreement into a false consensus. Separate evidence (what customers said) from any context (market) you pulled; never blend them.

Deliver

When you cite, use the inline linked-number badge per ${CLAUDE_PLUGIN_ROOT}/skills/using-evermuse/references/citations.md — 1. Citations are optional, but a voice-of-customer brief is far stronger with them, so prefer them here.

A brief that leads with the answer, then themes with quote blocks and inline citations:

**Short answer:** [1–2 sentences that directly answer the question.]

### Theme 1 — [name] ([N mentions / M accounts], mostly [sentiment])
> "[verbatim quote]" — [Name], [Meeting], [Date] [`1`](URL)
[one line of interpretation]

### Theme 2 — …

**Dissent / nuance:** [minority view, cited]

Save & hand off

Offer to save the brief: add_source(nature: "evidence", source_type: "document", tags: ["evermuse-plugin","research","voice-of-customer","<topic>"]). Then offer next steps: "Want a spec from this?" (/evermuse:spec), "Brainstorm solutions?" (/evermuse:brainstorm), or "Prioritize against the roadmap?"

Honesty when evidence is thin

If searches come back sparse, say so — "only 2 mentions, both from one account" — rather than dressing a weak signal as a trend. Thin evidence is itself a finding (maybe it's worth an interview: offer /evermuse:interview).

Further reading

• Research method draws on phuryn/pm-skills (MIT); grounding is Evermuse-native.