Research a product before you build it.
A study takes an idea through market research, user interviews and a strategy. You choose what to build from it, and those items arrive on the Builder board as new features.
A study is a set of research docs in your repo, shown on the board in Researcher mode. Unlike a feature's docs, research docs have no approval gate: you read them, run the interviews and decide what to hand off.
The same board, switched to research.
Each study is one row. The columns follow the research steps, and the last one shows what was handed to the builder.
Eight steps, one command.
Run any step with the command and the step name. With no step, it runs the next one, and it always stops after one step. A step waits for the ones it builds on; the brief is always open./specmanager:specmanager-research
research-loop runs 1 to 5, then stops
| 1 | Briefbrief | brief | Five questions (idea, market, audience, goals, constraints) turned into a short brief every later step reads |
| 2 | Market researchmarket | market | Market size, competitors and revenue, drafted as three parallel jobs and merged into one doc |
| 3 | Docsdocs · concepts | docs · concepts | The HTML Docs you pick: wireframes, a one-pager, a client or partner presentation, an investor deck, or your own type |
| 4 | User research planplan | plan | Who to interview, what to ask, and the decision rules the answers feed |
| 5 | Marketing planmarketing | marketing | Positioning, channels, a launch plan and the assumptions the interviews must confirm |
| — | Your interviews | You run them, then drop notes in the study's notes/ folder or paste them | |
| 6 | Findingsfindings | findings | What your interviews say, from your masked notes, with no invented scores or quotes |
| 7 | Strategystrategy | strategy | Learnings and a ranked build list (S1, S2, …), revising the market doc and marketing plan first |
| 8 | Handoffhandoff | handoff | The build-list items you pick become features on the Builder board |
new starts a study from a title and a one-line idea.
Steps 1 to 5 in one go
- Asks everything up front at most two rounds of questions, including which Docs you want.
- Drafts the five docs without stopping, then reviews them against each other and the brief.
- Stops before findings it approves nothing. Interviews are yours.
P3, ops lead at [redacted]: “We add people in batches every Monday, and nobody knows which role to give them.”
Pasted notes are saved already masked. Dropped .md, .txt and .pdf files are masked in place before findings reads them.
From a strategy item to a feature on the board.
Pick items from the strategy's build list. Each becomes a new feature with a small pill showing which study it came from, and an item already handed off is skipped. From there it's a normal feature, starting with its PRD.
- ✓S1 Team invites by email, in batches
- ·S2 Usage dashboard not picked yet
- ✓S3 Role presets admin, member, viewer
Research, answered
Is research a seventh stage?+
No. Research is its own board mode, with one row per study. Studies sit outside features, so they never touch a feature's gates, and research docs have no approval.
Does the research loop approve or build anything?+
No. It asks its questions up front, drafts the brief, market research, the Docs you picked, the user-research plan and the marketing plan, reviews them together, and approves nothing. It stops before findings, because findings need real interviews.
What happens to participant names in my notes?+
Pasted notes are saved already masked: a participant code such as P1 instead of the name, and [redacted] for emails, phones and addresses. Dropped .md, .txt and .pdf files are masked in place before findings reads them; other formats are skipped, and you're asked for a transcript.
Does research call a model or an API of its own?+
No. Research runs in your own Claude Code session, and every doc is written to .claude/specs/research/ in your repo. The plugin makes no model or API calls of its own.
What if the strategy changes after a handoff?+
The feature's origin pill switches to "strategy updated" with both versions. Its PRD is not marked stale, so review the strategy and update the PRD by hand if it matters.
Start a study on your next idea.
Research runs in your own Claude Code session and writes everything to your repo.