Back to skill

Security audit

Alfred Agent Governance

Security checks for vulnerabilities and agentic risk

Overview

This governance skill is not malicious, but it overstates important security controls and audit integrity that its own sample implementation does not provide.

Review this before installing as a security control. It appears to be documentation/sample code rather than an executable backdoor, but do not rely on it for approvals, rate limiting, resource limits, or tamper-evident audit logs until those controls are actually implemented and documented. Configure audit-log retention and access controls if used.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (3)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The feature list promises rate limiting, approval checks, and resource limits, but the provided enforcement code only checks deny_tools and deny_patterns before allowing execution. In a governance skill, missing enforcement of claimed controls can directly permit excessive tool use, unapproved privileged actions, or resource abuse while administrators wrongly believe protections are active.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill documents persistent audit logging of every tool call to disk but does not mention user notice, consent, retention, or safeguards around what operational metadata is stored. Even if only a params hash is logged, agent identity, tool usage, timestamps, and outcomes can expose sensitive behavioral data, and future implementations may expand logged fields without clear privacy boundaries.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation states that each audit log entry includes a signature, but the sample implementation writes only unhashed metadata and a truncated params hash with no authenticity or integrity field. This creates a false sense of tamper-evidence: operators may rely on logs for incident response or accountability even though an attacker with write access could alter or forge entries without detection.

Content

No source excerpt is available for this finding.