All skills
  • Claude Code skills
  • Product research
  • Open source

Product research kit

Six Claude Code skills for product research, from a competitor scan to a call on one feature. Every claim says where it came from.

I extracted them from the research stages of my own product-design pipeline. Each one works alone. Together they pass work to each other through one research/ folder: a skill that finds a sibling’s file reuses it instead of redoing the work.

Year
2026
Needs
Claude Code
Optional
Web search, a browser tool
License
MIT

How they work together

Each skill writes a file with a fixed name, so the next one finds it without asking. None of them requires another.

research/plus INDEX.md, one row per file
competitive-researchcompetitive-analysis.md
deep-researchdeep-research.md
ux-auditux-audit/<product>/audit_report.md
proto-personaspersonas.md
simulated-interviewinterview-<persona-slug>.md
feature-researchfeature-<slug>.md

Reads whatever of the above is in the folder: competitors, market research, audits, personas and interviews.

Arrows show what a skill picks up when it finds it. With an empty folder, it works from what you give it.

The honesty rules

This is the core of the kit. Facts come from you or from a source that was opened, and everything else is marked as what it is.

  • Same tag, same weightEvery claim that matters carries one tag, and a tag means the same thing in all six skills. An assumption in the personas stays an assumption in a feature report.
  • Nothing inventedNo invented numbers, quotes, reviews or URLs. When there is no data, the file says unknown and puts the item on its gaps list.
  • Synthetic is not evidenceA model playing a user drifts toward agreeing with you, which feels great and teaches nothing. What a persona says stays a hypothesis to check with real people.
[user]
Said by you in chat or written in your own files.
[source: Sn]
From a page or document that was actually opened. Sn is its row in the Sources list.
[snippet: Sn]
Seen only in a search-result excerpt, the page was not opened.
[unverified]
Known from model memory, not checked against a source.
[assumption]
Reasoning that was not checked. It points to an entry in the gaps list.
[estimate]
A size, effort or quantity guess, with what it assumes.
[synthetic]
Said by a simulated persona. A hypothesis, never evidence.

The six skills

Install one or all of them. Each lives in its own repository.

  • competitive-research

    Finds direct and indirect competitors, reads their pricing and flows, mines reviews and separates table stakes from differentiators.

    Writes research/competitive-analysis.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/competitive-research ~/.claude/skills/competitive-research

  • deep-research

    A sourced market, user and technology study on one product question, with the key figures checked against their sources.

    Writes research/deep-research.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/deep-research ~/.claude/skills/deep-research

  • ux-audit

    A heuristic audit by Nielsen’s 10 heuristics plus a cognitive walkthrough, from a live URL, screenshots or a local app.

    Writes research/ux-audit/<product>/audit_report.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/ux-audit ~/.claude/skills/ux-audit

  • proto-personas

    Personas built from the research you already have, every claim traced to a source or marked as an assumption, one of them primary.

    Writes research/personas.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/proto-personas ~/.claude/skills/proto-personas

  • simulated-interview

    Claude plays a persona in a jobs-to-be-done interview. You get a transcript and a debrief where every insight is synthetic.

    Writes research/interview-<persona-slug>.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/simulated-interview ~/.claude/skills/simulated-interview

  • feature-research

    Turns a feature idea into one report with a verdict: build it, build the smaller shape, validate first, or not now.

    Writes research/feature-<slug>.md

    GitHub (opens in a new tab)
    Install
    git clone https://github.com/VladPolyakov-art/feature-research ~/.claude/skills/feature-research

Why I built it

These six started as the research stages of my own product pipeline. I pulled them out so they run in any project, without the rest of the pipeline. What I kept is the method and the honesty rules, because that is the useful part: whoever reads the report can tell what is known from what is a bet.