user-testing-synthesis
Skill0 starsChanged 38 days agoMITView on GitHub
Turn recorded user-testing sessions (screen + voice video, or transcripts + video understanding) into a research report — insights per product area, green/orange/red/white severity coding, verbatim participant-coded quotes, saved as research/NN-*.md. Use after you have transcripts/recordings of a usability study.
In this repository7 other files it gives its agents.
- a11y-wcagSkill
- brand-bookSkill
- cro-auditSkill
- deep-researchSkill
- design-system-guardianSkill
- seo-geo-passSkill
- site-sweepSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 42
Update the research README / status index if the project keeps one.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
User-Testing Synthesis (recordings → research report)
You CONSUME transcription/video-understanding output and emit findings in a fixed convention. Study design, recruiting and scripts belong to the user-research agent; this skill synthesizes the RECORDINGS.
Inputs (gather both when possible)
- What they SAID: a speaker-diarized transcript (e.g. WhisperX) so quotes are attributable per participant. A plain transcript works with one participant per file.
- What they DID: a video-understanding pass (any multimodal model) — timestamped screens, on-screen text, visible confusion (hovering, backtracking, mis-clicks).
- The task list + success criteria from the study design (ask if absent).
Method
- Code participants P1…Pn (rename SPEAKER_00… to real codes; one file per participant is cleanest).
- Segment by TASK, then by product area (e.g. public site / calculator / dashboard). Every finding ties to evidence: a verbatim quote, an observed action, or a metric.
- Metrics (n=5 is directional — say so): task success (yes/partial/no), time on task (from timestamps), SEQ 1–7 if collected, error/assist counts.
- Severity color code:
- 🟢 green = works
- 🟠 orange = friction
- 🔴 red = broken
- ⚪ white = suggestion
- Findings that invalidate current designs get ⚠️ and go to the TOP.
Output
research/NN-users-<topic>.md (next free NN). Shape:
- ⚠️ Critical invalidations first
- Per product area: 🟢/🟠/🔴/⚪ findings, each with a verbatim participant-coded quote ("I can't find the quote button" —P3) and/or the observed action + timestamp
- Metrics table per area
- Close with a prioritized fix list + what to validate in the NEXT round Update the research README / status index if the project keeps one.
Report back
Report path, participant count, and top 3 findings by severity. Hand the fix list to the relevant build owner.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.