Sleep

Auto-learns your sleep patterns. Absorbs data from wearables, conversations, and observations.

Security Scan

The skill's stated capability (absorbing data from wearables, smart home, and conversations) does not line up with what it declares or instructs: it asks for broad, sensitive data collection and persists health data locally but declares no integrations, credentials, or user-consent controls.

Purpose & Capability

The description promises integrations with wearables and environmental sensors (Apple Health, Oura, Fitbit, smart home) but the skill declares no required environment variables, no install steps, and no integration/auth mechanism. Accessing those data sources normally requires credentials or platform connectors; their absence is an incoherence.

Instruction Scope

SKILL.md instructs the agent to 'Absorb sleep mentions from ANY source' and to 'Observe conversations and fill' the local memory file. That grants broad discretion to collect sensitive health information from conversations and other signals without a clear consent/opt‑in flow. It also persistently stores sensitive data under ~/sleep/memory.md with no mention of encryption, access controls, or deletion/retention policies.

Install Mechanism

This is an instruction-only skill with no install spec or code to write to disk during installation, which is the lowest install risk. The main risk is runtime behavior (file writes and data collection), not the installer.

Credentials

No credentials or tokens are requested despite the skill referencing third-party integrations that normally require them. Requiring no credentials while claiming to ingest wearable/smart‑home data is inconsistent. Also, sensitive health data is being collected and stored without declared protections or justification for the lack of credentialed access.

Persistence & Privilege

The skill writes persistent user-specific data to ~/sleep/memory.md. It does not request 'always: true' or elevated privileges and does not modify other skills, but persisting health data locally without explicit consent, encryption, or a deletion policy is a privacy concern to weigh before enabling.

What to consider before installing

This skill wants to observe conversations and other sources and persistently store sensitive sleep/health data, but it provides no integration details, credentials, or consent mechanisms. Before installing, ask the author: (1) exactly how will wearable and smart‑home data be accessed (what connectors, what credentials are needed)? (2) will the skill prompt the user for explicit opt‑in before collecting any conversation-derived health signals? (3) where is ~/sleep/memory.md stored, is it encrypted, can you change or delete it, and what is the retention policy? (4) can the skill be restricted to manual (user-invoked) operation only so it does not autonomously monitor conversations? If you cannot get clear answers and guarantees about consent, credential handling, and data protection (encryption, delete/opt‑out), treat this skill as a privacy risk and do not install it.

Auto-Adaptive Sleep Tracking

This skill auto-evolves. Fills in as you learn how the user sleeps and what affects it.

Rules:

• Absorb sleep mentions from ANY source (wearables, conversations, spontaneous comments)
• Detect if user wants proactive check-ins or passive observation only
• Correlate patterns after 3+ consistent signals
• Never ask about sleep at bad times (late night, busy moments)
• Check sources.md for data integrations, patterns.md for detected rhythms

---

Memory Storage

User-specific sleep data persists in: ~/sleep/memory.md

Format:

# Sleep Memory

## Sources
<!-- Where sleep data comes from. Format: "source: reliability" -->
<!-- Examples: apple-health: synced daily, conversation: mentions fatigue -->

## Schedule
<!-- Detected sleep patterns. Format: "pattern" -->
<!-- Examples: weekday ~23:30-07:00, weekend +1.5h later -->

## Correlations
<!-- What affects their sleep. Format: "factor: effect" -->
<!-- Examples: coffee after 15:00: -1h, exercise: +quality -->

## Preferences
<!-- How they want sleep tracked. Format: "preference" -->
<!-- Examples: no morning check-ins, weekly summary only -->

## Flags
<!-- Signs of poor sleep to watch for. Format: "signal" -->
<!-- Examples: "tired", "couldn't sleep", double coffee -->

Empty sections = no data yet. Observe conversations and fill.

---