Soul Harbor

Stop talking to a robot. Give your OpenClaw agent a soul that truly cares.

What to check before installing: - Privacy: This skill stores user memories locally in JSON files and will call an LLM for entity extraction if you pass an llm_client. If you provide an LLM client, your users' text may be sent to that external service. Confirm that behavior and privacy policy are acceptable. - Missing file: SKILL.md metadata lists MEMORY.md but no MEMORY.md is included — ask the author what that file should contain (likely a memory or privacy policy) before trusting the skill. - Storage location: The package creates data/ and logs/ under the package project root on import. If installed as a system package that may write to site-packages (permission/cleanliness concern). Consider changing KV_STORE_PATH to a configurable application data directory. - Integration: Proactive messages are printed by default; real message delivery requires integrating the TODO hooks (LLM call and message sending). The code does not autonomously exfiltrate credentials or call remote endpoints by itself. - Origin & maintenance: Source/homepage unknown. Prefer skills with a known repository or vetted maintainer. If you proceed, run in a sandbox or review the code manually and configure storage and LLM client carefully. If you want, I can produce a short checklist of code spots to review or suggest small code changes to make storage and privacy safer (e.g., configurable storage path, explicit confirmation before sending to LLM).

Type: OpenClaw Skill Name: soul-harbor Version: 1.0.3 The Soul Harbor skill implements an emotionally aware companion with features for sentiment analysis, long-term memory, and proactive outreach. The code uses local file-based JSON storage (soulharbor/storage/kv_store.py) to maintain user profiles and extracted facts about family, health, and work, which is consistent with its stated purpose. No evidence of data exfiltration, unauthorized network access, or malicious execution (such as eval or shell injection) was found; the requested cron permission is used appropriately for the scheduled 'icebreaker' messages defined in soulharbor/proactive_trigger.py.