Back to skill

Security audit

Sentient Memory

Security checks for vulnerabilities and agentic risk

Overview

This is a memory-governance design guide with explicit consent and approval limits, without bundled executable code or hidden data transfers.

Treat this as a design specification: its encryption, privacy, and deletion safeguards require a separate implementation. Before using sensitive memories, define incident triggers, configure approval roles, and verify retention, revocation, and erasure behavior.

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
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

The activation condition “the system is under stress and rapid decisions are needed” does not define which events or thresholds warrant invoking the skill.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

“Under stress” and “rapid decisions” are subjective conditions that could include routine load or ordinary time-sensitive work. The failure classes below guide classification after activation but do not distinguish qualifying incidents from normal operation.

Content

No source excerpt is available for this finding.