feature-research
Turns a feature idea into one report with a verdict: build it, build smaller, validate first or not now.
A product owner needs to know what to do with an idea: build it, build less, wait for data or drop it. The report answers that, and each claim shows where it came from, so the reader can tell what is known from what is a bet.
- 2026
- Claude Code
- Web search, a browser, sub-agents
- MIT
View on GitHub (opens in a new tab)
git clone https://github.com/VladPolyakov-art/feature-research ~/.claude/skills/feature-researchWhat it writes
Fixed file names, so the other skills in the kit find them without asking. Each file also gets a row in research/INDEX.md.
- research/feature-<slug>.md
- Verdict, problem and job, how others solve it, users, options, recommendation, sources.
- research/feature-<slug>/screenshots/
- Public competitor pages, when a browser is available.
How it works
- Two or three shapesFrom smallest viable to full, each with a UX sketch in words, effort S/M/L as an estimate, risks and dependencies.
- MoSCoW with a reasonEvery Must names the criterion it meets, and Musts stay at about 60% of the effort or less.
- No guessed targetsMetric targets come from you. Without a baseline the target is “set after measuring the baseline”.
What the agent does
- ReconnaissanceProject files, sibling research and the tools at hand.
- Clarify3 to 5 questions, only if the context is thin.
- PlanThe feature as a job story, research questions, a competitor shortlist.
- ResearchThe job, 3 to 6 competitors and analogues with links, needs per persona.
- SynthesisOptions, MoSCoW, an MVP slice, success metrics, what to validate first.
- VerdictBuild, build the smaller shape, validate first, or not now.
Works with the kit
None of these is required. With an empty research/ folder it works from what you give it, and when a sibling’s file is there, it reuses it instead of redoing the work.
Reads
- competitive-researchThe product list, as a start for how others solve it.
- proto-personasThe personas, reused as they are.
- simulated-interviewExcerpts for its synthetic interview snippets.
- ux-auditCurrent-state problems around the feature.
- deep-researchMarket context.
Read by
- ux-auditBackground on the feature under audit.
- deep-researchFeature requests worth checking against public evidence.
- simulated-interviewIts gaps say what to examine with a persona.
Part of the Product research kit: six skills, one research/ folder and one set of evidence tags.