Back to skill

Security audit

ML Engineering

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent ML engineering playbook with templates and checklists, not a skill that secretly runs code or accesses sensitive data.

This skill appears safe to install as a reference playbook. Users should still review any generated deployment, monitoring, or retraining configuration before applying it to real infrastructure, especially because some invocation phrases are broad and may need clarification in mixed-skill environments.

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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • 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
Findings (4)

Chaining Abuse

High
Category
Tool Misuse
Content
COPY src/ ./src/

# Non-root user
RUN useradd -m appuser && chown -R appuser /app
USER appuser

# Health check
Confidence
75% confidence
Finding
Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The README presents "Deploy model" as a standalone invocation phrase, but it is generic and likely to overlap with common requests in many contexts. The description does not provide trigger constraints, activation boundaries, or negative examples to clarify when this skill should activate versus other deployment-related skills.

Vague Triggers

Medium
Confidence
89% confidence
Finding
"Set up monitoring" is a broad phrase that could refer to infrastructure, application, security, or ML monitoring, making accidental invocation more likely. The README does not define explicit trigger scope or exclusions to distinguish this skill from other monitoring-related skills.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The skill defines very broad natural-language trigger phrases such as "Select model", "Evaluate model", and "Deploy model" without any namespacing or disambiguation. In an agent environment, these generic phrases can be invoked accidentally by ordinary user requests or collide with other skills, causing unintended execution and making prompt-routing easier to manipulate.

Static analysis

No suspicious patterns detected.