ClawHub
Back to skill

Security audit

Agent Memory Design

Security checks for vulnerabilities and agentic risk

Overview

This memory-design skill is coherent, but it encourages broad long-term storage of conversations, profile details, documents, emails, and session state without privacy controls.

Review this skill carefully before installing. It is not executing code, but it teaches agents to build durable memory stores. Use it only with explicit policies for what may be remembered, what must never be saved, how users opt in or delete data, how long records are retained, and how private files, messages, credentials, and regulated data are excluded or protected.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly encourages building persistent memory and storing user/profile/session information, but it does not include guardrails for consent, data minimization, sensitive-data filtering, retention limits, or access controls. In a memory-design skill, this omission is security-relevant because it normalizes writing potentially sensitive user data to durable storage by default, which can lead to privacy violations or later data exposure.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This section promotes broad retention of past conversations, decisions, facts, documents, emails, and meeting notes across searchable and archival tiers without sufficiently constraining scope or sensitivity. That creates a real risk of over-collection and resurfacing of confidential or personal data beyond what is necessary for the agent's function, especially in long-running deployments.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The write-trigger guidance tells the agent to persist newly learned facts, decisions, task status, and pre-compaction context as a default operational behavior. Without filtering or approval gates, this can cause routine persistence of sensitive conversation details, mistaken inferences, or transient operational data that users did not expect to be stored long-term.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The compaction checklist and recovery steps direct saving and later re-reading conversation topics, verified facts, task state, and pending decisions to survive context loss. In context, that increases the chance that sensitive user data or confidential operational details are retained and resurfaced after the user expects them to have fallen out of context, expanding exposure and persistence risk.

Content

No source excerpt is available for this finding.