Back to skill

Security audit

Engramai

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent local AI memory helper, but users should be careful about package pinning, stored personal data, and automated memory deletion.

Install in a virtual environment, pin and review the engramai package version where possible, keep the memory database protected, avoid storing secrets or sensitive personal data unless explicitly intended, and do not schedule automatic forget operations without backups or clear retention thresholds.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:15
Finding

Unpinned Third-Party Package Installation

Content
View full analysis
Remediation
View remediation
``` 2. Distribute a hash-locked requirements file and require hash verification: ```bash pip install --require-hashes -r requirements.txt ``` 3. Record hashes for every package artifact and transitive dependency used by the installation. 4. Review the source and published artifact for the pinned release, including build-system configuration and installation hooks. 5. Use a trusted package index and explicitly configure the expected index to reduce dependency-confusion and repository substitution risks. 6. Install and execute the package in an isolated virtual environment or container under a dedicated, least-privileged account. 7. Restrict filesystem, credential, environment-variable, and network access for the MCP server and maintenance processes to only what memory management requires. 8. Adopt a controlled update process in which dependency upgrades are reviewed, tested, scanned, and assigned new integrity hashes before deployment. 9. Update both the metadata declaration and installation example so automated installers and human operators use the same pinned, verified release. ]]>
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The documentation promotes consolidation, pruning, and forgetting operations that can permanently modify or delete stored memories, but it does not warn users about irreversibility, backup needs, or review steps before deletion. In a memory system used by agents, this can cause accidental loss of important records, undermine auditability, and make recovery impossible if the commands are triggered automatically via cron or heartbeat.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill explicitly recommends storing user preferences, habits, and other personal information in a local memory database, but it does not include a clear privacy notice, consent guidance, retention policy, or safeguards around sensitive data handling. In an AI-agent memory context, this can lead to silent collection and long-term persistence of personal data, increasing privacy, compliance, and misuse risk even if the storage is only local.

Content

No source excerpt is available for this finding.