Crisis Detector
WarnAudited by ClawScan on May 10, 2026.
Overview
The skill is for a coherent safety purpose, but it describes automatic crisis monitoring and contacting emergency contacts or authorities without enough clear user-control, consent, or data-retention safeguards.
Only use this skill if you can configure strict safeguards: no automatic authority or emergency-contact notifications without human review, explicit consent for any monitoring, clear retention/deletion rules for mental-health data, and verification of the referenced npm package before installation.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A mistaken classification could disclose highly sensitive mental-health information or trigger emergency/law-enforcement involvement without the user's clear control.
The documented default escalation path includes contacting authorities and emergency contacts based on automated crisis classification, but the artifact does not define mandatory human review, confirmation, contact scoping, or safeguards against false positives.
"onImminent": ["emergency_intervention", "contact_authorities", "notify_emergency_contacts", "continuous_monitoring"]
Require explicit user consent where appropriate, human review before external escalation, preconfigured verified contacts, jurisdiction-aware emergency guidance, and safe defaults that show resources rather than contacting third parties automatically.
Sensitive mental-health data could be stored or reused in future decisions in ways the user does not understand, increasing privacy risk and the chance of incorrect escalations.
The skill proposes retaining or reusing sensitive crisis and mental-health context over time, but the artifact does not specify storage location, retention period, deletion controls, access controls, or how historical signals are prevented from poisoning later decisions.
"Historical Analysis": "Track user patterns over time" and "historicalContext": true, "trackPatterns": true
Make historical tracking opt-in, minimize stored data, define retention and deletion controls, encrypt or otherwise protect any logs, and separate resource suggestions from persistent risk scoring unless the user has clearly agreed.
The agent could continue observing or acting on sensitive user activity after the immediate interaction, creating privacy and autonomy concerns.
The skill describes ongoing monitoring that can persist beyond a single user request, but it does not define when monitoring starts, who can stop it, how long it runs, or what data sources it watches.
"Real-time Monitoring": "Continuous monitoring of at-risk users" and actions including "enable_monitoring", "enable_intensive_monitoring", "continuous_monitoring"
Require explicit opt-in for monitoring, visible status indicators, a simple stop command, time limits, clear data-source boundaries, and no background monitoring by default.
Users cannot verify from these artifacts what the referenced package actually does, especially important for a high-stakes mental-health workflow.
The skill documentation points to an external npm package, while the provided review context contains no code files or install specification to inspect that package's behavior.
Installation: npm; Package: `@raghulpasupathi/crisis-detector`; `npm install @raghulpasupathi/crisis-detector`
Review the npm package source, pin versions, provide an auditable install spec or lockfile, and document exactly what code runs before relying on it.