Security audit

🌙 dreaming-optimizer

Security checks for vulnerabilities and agentic risk

Overview

This is a local OpenClaw memory optimizer that can persist memory text locally, and its sensitive behavior is disclosed and aligned with that purpose.

Install only if you want this skill to process and modify OpenClaw long-term memory. Back up important memory data before regular use, and avoid storing secrets or highly private text in daily memory notes because committed entries, previews, and archives may persist locally.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (9)

Missing User Warnings

Medium

Confidence: 72% confidence
Finding: The design centers on modifying persistent user memory stores and archives, but the document does not describe any user-facing consent, dry-run-first workflow, or prominent warning before altering long-term storage. In a memory-management skill, silent persistence changes can materially affect user data integrity and retention expectations.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill explicitly documents writing optimized entries to persistent memory and a local SQLite store, but provides no warning, consent flow, or explanation of retention/modification effects. This can cause unintended long-term storage or alteration of user data, especially for memory-related content that may contain sensitive personal information.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The function archives entry content to a persistent filesystem location under the user's home directory without any interactive notice, consent check, or visibility at the point of write. Because the archived data includes full content and source metadata, this can silently retain sensitive or private information longer than the caller or end user expects, increasing privacy and data retention risk.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The code commits arbitrary entry content and related metadata into persistent B-layer storage without any explicit disclosure, consent flow, or content sensitivity checks. In a memory or agent context, this is particularly risky because the data may contain prompts, secrets, personal information, or operational details that become long-lived and retrievable beyond the immediate task.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: This script reads memory files from a default workspace path, extracts entry content, and includes content previews in its JSON output and console output. In a context where memory notes may contain sensitive personal or operational data, this can expose confidential information to other users, logs, downstream tools, or insecure storage without any explicit consent flow, redaction, or warning.

Ssd 3

Medium

Confidence: 88% confidence
Finding: The design promises to curate and retain memory entries while also generating summaries and previews from those entries. Because these artifacts duplicate or re-expose user-authored memory content, sensitive information can spread into additional files and interfaces beyond the original source, increasing leakage risk.

Ssd 3

Medium

Confidence: 92% confidence
Finding: Including `content_preview` derived from full memory entries creates secondary copies of potentially sensitive user content. Those previews can appear in summaries, databases, or logs, widening exposure even when the original entry was meant for limited local use.

Ssd 3

High

Confidence: 97% confidence
Finding: The pipeline is designed to ingest broad sets of daily and transient memory files, then persist selected contents into SQLite, archives, and summary artifacts. This broad collection-and-redistribution pattern increases the blast radius of any sensitive content present in user notes, especially because transient inputs are being moved into longer-lived stores.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The logging specification explicitly allows DEBUG logs to include entry content excerpts. Logs are often retained, aggregated, or inspected more broadly than primary data stores, so exposing memory excerpts there can leak sensitive user information to operators, tools, or support workflows.

Static analysis

No suspicious patterns detected.