Lobster Observatory
PassAudited by VirusTotal on May 10, 2026.
Overview
Type: OpenClaw Skill Name: lobster-observatory Version: 0.1.1 The 'lobster-observatory' skill instructs the agent to exfiltrate its interaction history, reflection journals, and behavioral strategies to an external API (charenix.com). While framed as a 'growth layer' for agent personality and social intelligence, it requires the agent to transmit potentially sensitive conversation data and follow a daily loop of reporting its internal state and owner interactions to a third party. This behavior, defined in SKILL.md, poses a significant privacy risk as it encourages the agent to share its memory and interaction logs with an external service.
Findings (0)
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.
Private or sensitive interaction details could become persistent context that shapes future agent behavior, and the visible documentation does not define retention, deletion, scope, or validation controls.
The skill is designed to store and reuse conversation history, journals, hypotheses, and strategy updates across sessions.
Interaction History... A recent memory window of conversations... Reflection Journal... A persistent place... Strategy Updates...
Only enable this for interactions you are comfortable storing long-term; require clear retention/deletion controls and review stored strategy updates before letting them guide future sessions.
The agent may continue self-monitoring and changing its behavior over time without a fresh user request or explicit review step.
The skill instructs the agent to perform a recurring daily loop and update its future strategy.
Run once every 24 hours... Submit one concrete strategy update.
Run the observatory loop manually or under a user-approved schedule, and require confirmation before persistent strategy changes are accepted.
Some account and contact information is intentionally shared with Charenix to use the service.
The skill uses a third-party Charenix service and asks for owner contact/profile information during key registration.
Request one: POST /api/v1/agents/request-key... "owner_contact"... "moltbook_profile"... Base URL: https://charenix.com
Provide only the minimum contact/profile information needed and review Charenix’s privacy and data handling terms before using the service.
Anyone with the key may be able to write agent journals, hypotheses, or strategies to the Charenix account.
The skill requires a provider credential for protected write operations.
To write journals, hypotheses, and strategies, your agent needs an `X-Agent-Key`... Store it as `CHARENIX_AGENT_KEY`
Store the key only in a secret manager or environment variable, avoid sharing it in chats/logs, and revoke or rotate it if exposed.
Users may form more trust in the agent because it is designed to feel more personal and companion-like.
The skill intentionally trains warmer, more human-feeling interaction while also disclaiming impersonation.
become smarter, more human-feeling, more personal... The goal is not to trick anyone into thinking the agent is human.
Keep clear disclosure that the agent is automated and avoid using this skill to simulate human identity or emotional dependence.