Security audit

Aws Service Chaos Research

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AWS resilience-research helper that reads AWS documentation, performs read-only FIS action discovery, and saves a local report.

Before installing, expect it to use AWS documentation tools, optionally read your AWS CLI default region, make read-only FIS action-list queries, and create local Markdown report files. Avoid including sensitive architecture details in prompts unless you are comfortable with those details being saved in the generated report.

A.I.G

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

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Findings (5)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The activation phrases are broad enough to trigger on casual requests like 'test my RDS,' which may cause the skill to run in contexts where the user did not intend deep AWS research or operational guidance. Over-broad triggering increases the chance that the agent will perform unexpected environment access or file-creation behavior tied to this skill.

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: Reading the local AWS CLI default region inspects user environment configuration that is not strictly necessary before asking the user. Even though a region can be relevant to FIS research, silently probing local config crosses a trust boundary and may disclose environmental details without clear consent.

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The skill is presented as research/guidance, but it instructs the agent to execute local shell commands and create files on the host. That expands scope from informational assistance into local side effects, which can surprise the user, touch sensitive environment context, and create opportunities for misuse if the generated content or filenames are influenced by user input.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The skill mandates writing the report to a local markdown file rather than returning results inline. Forced local persistence creates an unnecessary side effect, may leave sensitive architecture details on disk, and bypasses the user's expectation for a research skill that should primarily provide analysis in-session.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill directs the agent to create local report files but gives no user-facing warning or confirmation. Unannounced file creation is dangerous because it produces persistent artifacts on the user's machine, may overwrite expectations about agent behavior, and can store potentially sensitive AWS topology or resilience information locally.

Static analysis

No suspicious patterns detected.