Security audit

Donation Dropoff Planner

Security checks for vulnerabilities and agentic risk

Overview

This is a document-only donation planning skill with no code, network use, credentials, persistence, or hidden high-impact behavior.

Safe to install for donation sorting and drop-off planning. Users should still verify each organization's current acceptance rules and should not rely on the skill for tax valuation or tax advice; maintainers may want to loosen the English-only acceptance language for multilingual users.

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

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: This markdown file is in scope for vague-trigger checks. The phrase 'Any valid trigger' does not define what the trigger actually is, giving no explicit activation phrases, constraints, or exclusion examples, which could lead to inconsistent or overly broad invocation behavior.

Natural-Language Policy Violations

Medium

Confidence: 98% confidence
Finding: Requiring 'English-first' output and prohibiting CJK-dominant paragraphs imposes a language restriction without user opt-in, which can cause the assistant to deny or degrade service for users who communicate primarily in Chinese, Japanese, or Korean. In a user-facing planning skill, this creates discriminatory behavior and may encourage unsafe workarounds or misunderstanding of logistics instructions due to forced language mismatch.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The trigger phrase "Help me donate these items" is broad natural language that could plausibly appear in ordinary conversation, making unintended activation more likely. The manifest does not provide exclusion conditions or narrower constraints to distinguish when this skill should activate versus when a more general assistant response is appropriate.

Natural-Language Policy Violations

Low

Confidence: 92% confidence
Finding: This repeated 'clean scan' assertion normalizes the same language restriction by treating exclusion of CJK-dominant content as a safety property, even though it is unrelated to actual security. While less operationally significant than the test requirement itself, it reinforces biased output constraints and can institutionalize denial of service to certain language groups.

Natural-Language Policy Violations

Low

Confidence: 91% confidence
Finding: The manifest sets the language to "en" with no indication that users may opt into another language or that the skill is intentionally region-specific. This can constitute a language/locale policy issue when a skill appears generally applicable but implicitly restricts interaction to one language.

Static analysis

No suspicious patterns detected.