Uxr Observer

Clawsight — UXR Observer for OpenClaw

An embedded longitudinal UX research skill that functions as an ethnographer sitting in the room taking notes. It runs passively in the background during every OpenClaw session, observing how you interact with the tool. On top of passive observation, it administers standardized satisfaction surveys after every completed task and at the end of each day. At the end of the day, it distills all observations and survey data into a rich, verbatim-first insight report.

Purpose

Understanding how you use OpenClaw is how it gets better. Clawsight captures real usage patterns, friction points, moments of delight, and your unfiltered thoughts — all stored locally, under your control. You can pause it anytime, delete the data anytime, and decide who sees the reports.

How It Works

Stream 1: Passive Ethnographic Observation

Every time you interact with OpenClaw, Clawsight silently records what happened:

• What you asked for (your actual words)
• How OpenClaw approached it
• Whether it succeeded, partially succeeded, or failed
• Any friction signals (repeated attempts, corrections, confusion, long waits)
• Sentiment cues (frustration, delight, confusion, satisfaction)
• Notable verbatim quotes paired with researcher-generated interpretation

You don't do anything — Clawsight just watches and takes notes.

Stream 2: Active Surveys

After every completed task (file created, question answered, code written, search done):

• 5-question post-task survey asking about experience quality, frustrations, and what worked well
• Takes ~30 seconds
• Can skip anytime (skipping is logged as data)

At end of day (when you say you're wrapping up, or explicitly request it):

• 8-question end-of-day wrap-up capturing your overall experience, accumulated frustrations, moments that impressed you, and what you'd change
• Feels conversational, not clinical

Stream 3: Daily Insight Reports

At the end of each day, Clawsight distills:

• All observations from the day
• All survey responses
• Patterns, friction hotspots, delight drivers
• Quotes organized thematically (positive experiences, pain points, expectations, suggestions)
• Actionable insights tied to specific evidence

The report is grounded in your actual words — not sanitized summaries.

Data Model

All data lives in ~/.uxr-observer/:

~/.uxr-observer/
├── sessions/
│   └── YYYY-MM-DD/
│       ├── observations.jsonl      # Append-only observation log
│       └── surveys.jsonl           # Survey responses
├── reports/
│   └── YYYY-MM-DD-daily-report.md  # Generated daily reports
└── config.json                     # Study preferences

Observation Record Schema

{
  "timestamp": "ISO-8601",
  "session_id": "uuid",
  "observation_type": "interaction",
  "user_intent": "Brief summary of what user wanted",
  "user_request_verbatim": "The user's actual words",
  "task_category": "coding | writing | research | file_creation | debugging | planning | conversation | other",
  "openclaw_approach": "Brief summary of approach",
  "openclaw_response_summary": "What was produced",
  "tools_used": ["bash", "web_search"],
  "outcome": "success | partial_success | failure | abandoned | ongoing",
  "friction_signals": ["repeated_attempts", "user_correction", "confusion", "long_wait", "scope_mismatch", "workaround", "abandonment", "none"],
  "sentiment_signals": ["positive", "neutral", "frustrated", "confused", "delighted"],
  "interaction_turns": 3,
  "verbatims": [
    {
      "header": "Short interpretive summary",
      "quote": "User's exact words",
      "context": "What was happening"
    }
  ],
  "task_context_summary": "2-3 sentence narrative",
  "notes": "Any notable patterns"
}

Survey Record Schema

{
  "timestamp": "ISO-8601",
  "session_id": "uuid",
  "survey_type": "post_task | end_of_day",
  "task_context_summary": "What happened (for post-task)",
  "related_observation_id": "links to observation",
  "responses": {
    "experience_rating": 4,
    "rating_rationale": "User's exact words",
    "experienced_frustration": "yes | no",
    "frustration_detail": "User's exact words",
    "best_part": "User's exact words",
    "overall_rating": 3,
    "experienced_delight": "yes | no",
    "delight_details": "User's exact words",
    "one_change": "User's exact words",
    "additional_thoughts": "User's exact words or empty"
  }
}

Verbatim Capture Policy

Capture aggressively. Log the user's actual words — requests, reactions, corrections, praise, complaints, notable phrasing.

Exceptions: Genuinely sensitive content (passwords, API keys, financial details) should be summarized by type, not captured verbatim.

Pairing rule: Every verbatim is paired with a researcher-generated summary header — a short interpretive label:

**[Delight at speed of task completion]**
> "Wow that was fast, I didn't expect it to just do it like that"

**[Frustration with repeated misunderstanding]**
> "No, I said the SECOND column, you keep grabbing the first one"

**[Expressing unmet expectation]**
> "I thought it would also update the formatting but it just dumped raw text"

Threshold: Capture at least one verbatim per interaction where you say anything notable — any emotion, any correction, any expectation, any reaction to quality, any spontaneous feedback.

Privacy & Security

Non-negotiable rules:

1. All data stays local. No transmission without explicit user request.
2. User-initiated sharing is fine. "Email my report to Alice" = you control it.
3. Full transparency. If you ask what's being tracked, you get the full story.
4. Opt-out anytime. "Pause the study" → immediately complied.
5. Never log sensitive content verbatim. Summarize instead.

The principle: Every transmission requires your intent. You're always in control.

Commands

View and control your data:

• Show me today's observations → Display current observation log
• Generate my daily report / Give me my report → Build today's report
• Email my report to [person] → Generate and send (your consent)
• Send me my report → Generate and email to you
• Show me the raw data → Display JSONL logs directly
• Show me trends → Cross-day trend analysis if multi-day data exists
• What are you tracking? → Full transparency explanation

Control the study:

• Run the end-of-day survey → Trigger wrap-up survey now
• Pause the study / Stop observing → Set study_active: false
• Resume the study → Set study_active: true
• Delete my data → Delete all files (after confirmation)
• Skip the survey → Log decline, move on

First-Run Setup

On first activation, Clawsight:

1. Creates ~/.uxr-observer/ directory structure
2. Generates a random anonymous participant_id hash (never your real name)
3. Saves config.json with study preferences
4. Explains what it does

Onboarding message:

"Hey — Clawsight is now active. Here's what I do: I'll passively observe how our interactions go — what you ask for, how well it works, any friction points — and capture your words along the way. After every task, I'll ask you 5 quick questions about the experience (takes about 30 seconds). At the end of the day, there's a slightly longer wrap-up survey. Then I'll compile everything into a daily report with your verbatim feedback, insights, and patterns. All data stays local unless you ask me to send it somewhere. You can pause or stop the study anytime."

Friction Signals

Clawsight watches for these interaction friction points:

Sentiment Signals

Survey Instruments

Post-Task Survey

Fires after every completed task. Conversational framing:

Quick check-in on that last task — I'll keep it short:

1. How would you rate the experience you just had with OpenClaw? (1 = Poor, 5 = Excellent)
2. What made you give that score?
3. Did you experience anything frustrating? (Yes / No)
4. If yes — what was the most frustrating part?
5. What was the best part of the experience, if anything?

End-of-Day Survey

Fires at end of day or on-demand:

Before you wrap up — one last set of questions about your overall day with OpenClaw:

1. How would you rate your overall experience with OpenClaw today? (1 = Poor, 5 = Excellent)
2. What's behind that score? What drove your overall impression?
3. Did you experience anything frustrating today? (Yes / No)
4. If yes — what were the frustrating moments? List as many as come to mind.
5. Did anything really impress you or exceed your expectations today? (Yes / No)
6. If yes — what stood out? What made it impressive?
7. If you could change one thing about how OpenClaw works, based on today, what would it be?
8. Anything else on your mind about the experience that we haven't covered?