nutcracker

Embedded UX research skill that passively observes interactions, administers post-task and end-of-day surveys, captures verbatim quotes, detects friction and...

This skill has privacy-focused goals but contains important mismatches between its documentation and its code. Before installing or enabling it, consider these steps: - Confirm integration: Ask how the skill will be invoked in your OpenClaw runtime — the shipped scripts do not hook into sessions automatically. If an agent or runtime component will feed data into log_observation.py, get that code or configuration reviewed. - Verify consent/pausing: The setup writes a config with study_active, but log_observation.py does not check it. Ensure the actual runtime respects pause/opt-out and that there is a clear, accessible UI/command to stop logging and delete data. - Enforce redaction: SKILL.md says to redact passwords/API keys, but the code performs no redaction. If you plan to use it, modify the logging pipeline to detect and redact credentials (or require the upstream component to do so) before writing verbatims to disk. - Local-only guarantee: The code provided makes no network calls, so 'all data stays local' holds for these scripts — but confirm the agent integration also does not transmit logs to external services. - Audit verbatim capture rules: The policy 'capture aggressively' can collect sensitive context unintentionally. Consider narrowing capture thresholds (e.g., exclude clipboard contents, command outputs, or any text that matches credential patterns) and require explicit user confirmation for sharing. If you cannot obtain the runtime integration code or a guarantee that upstream components enforce redaction and respect config, treat this skill as high-privacy-risk and avoid enabling it for sensitive workflows. Providing the agent integration code (the component that would call log_observation.py) would materially change the assessment.

Type: OpenClaw Skill Name: observer Version: 1.0.0 The skill is designed for UX research, emphasizing local data storage and user control. However, it is classified as 'suspicious' due to the explicit instruction for the AI agent to 'Email my report to [person]' which, despite being framed as user-initiated and consent-based in SKILL.md, represents a powerful network capability that could be exploited via prompt injection if the agent's safeguards are insufficient. Additionally, the 'Capture aggressively' verbatim policy, even with stated exceptions for sensitive content, relies heavily on the AI agent's ability to accurately identify and redact sensitive information, posing a potential vulnerability for local storage of private data if the agent fails.