Back to skill

Security audit

ADK Skill Patterns

Security checks for vulnerabilities and agentic risk

Overview

This Markdown-only skill provides disclosed templates and guidance for designing agent skills, with no hidden execution, persistence, or data access behavior found.

This skill is reasonable to install as a pattern reference. If you reuse its templates to publish new skills, narrow the description fields to the exact domain and activation conditions so they do not trigger on unrelated user requests.

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 (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill description includes broad trigger phrases like 'write, create, or draft a report, summary, or analysis document,' which can match a wide range of ordinary user requests and cause over-selection of this skill. In an agent system, overly broad activation can route unrelated tasks into a templated report workflow, increasing the chance of inappropriate file loading, unintended context use, or unreliable behavior across many benign prompts.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger phrases include broad terms such as "help me plan" and "design a system," which can match many unrelated user requests and cause the skill to activate when the user did not intend to invoke a structured project-planning workflow. This can lead to incorrect routing, unnecessary collection of user inputs, or interference with more appropriate skills, though it does not directly introduce code execution or data exfiltration risk.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill description is broadly scoped to trigger on common requests like 'feedback' or 'code audit', which can cause the reviewer skill to activate in situations beyond narrowly intended Python code review. Over-broad activation increases the chance of misrouting user requests, unnecessary context access, or accidental invocation in mixed-task conversations where a more appropriate or safer skill should handle the request.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest-style description says to use the skill when the user asks to 'document a module, generate API docs, or create documentation from code,' which is fairly broad and lacks explicit boundaries or exclusion conditions. In a markdown/manifest context, this can cause the skill to match a wide range of ordinary requests without clarifying when the pipeline should not activate.

Content

No source excerpt is available for this finding.