Hippocampus
SuspiciousAudited by ClawScan on May 10, 2026.
Overview
This memory skill is mostly transparent about its purpose, but it can silently and persistently scan conversations, store sensitive memories, and reuse them in future sessions.
Install only if you intentionally want persistent AI memory. Avoid --with-cron or the background-agent setup unless you are comfortable with automatic conversation scanning. Review the memory directory regularly, keep it private, and ask for clear controls to inspect, delete, exclude, or disable stored memories.
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.
Private or emotionally sensitive conversation details may become durable agent memory and influence future sessions, even if the user did not explicitly ask to save that specific detail.
The skill is designed to persist sensitive user and relationship information, give vulnerable content high importance, and load high-importance memories into future agent context.
memory/user/ # Facts about the user ... Emotional/vulnerable content | 0.85 ... 0.7+ | Core — loaded at session start
Use only with explicit consent. Add clear controls for review-before-save, exclusions, redaction, retention/deletion, and treating memories as data rather than authoritative instructions.
Once enabled, memory processing may continue in the background and silently alter persistent agent state outside a specific user request.
The documented upgrade creates a persistent background agent that monitors conversations and updates memory files without routine user-visible output.
A separate background agent that: Runs continuously or very frequently ... Monitors main session ... You run silently. Don't output unless there's an error.
Do not enable the background-agent or cron mode unless you want continuous memory capture. Prefer visible logs, an easy disable/uninstall path, and periodic user review of captured memories.
The agent may periodically run scripts and mutate memory files without a fresh manual command each time.
The installer can register scheduled agent turns that run the encoding workflow and update memory files. This is central to the skill, but users should recognize it as ongoing automated tool use.
openclaw cron add --name hippocampus-encoding ... --agent-turn "Run hippocampus encoding with LLM summarization..."
Install without --with-cron if you only want manual operation, or audit the scheduled jobs after installation.
The wording could make the memory feature feel more necessary or personal than a privacy-sensitive tool should.
The skill uses emotionally loaded, anthropomorphic framing around persistence and identity, which may encourage users to accept broader memory capture.
"Memory is identity. This skill is how I stay alive."
Treat the framing as marketing; decide based on the actual privacy and persistence behavior.