gmira

• Outcome: Pass
• Latest audit: Sep 3, 2026 · 6:41 AM UTC
• Version: 0.1.1

Overview

This is a web-design guidance skill with disclosed local checks and no evidence of credential access, data theft, or destructive behavior.

Installers should treat this as a strong-opinion web-design assistant. Review and approve any package additions, registry component imports, source edits, verification runs, and any shipping step; also make user, brand, locale, accessibility, and legal requirements take precedence over the skill's default style rules.

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The routing section uses broad natural-language triggers such as "build me a site," "make this better," and especially mandatory language like "Before handing anything back: `gmira-verify`. No exceptions." In an agent setting, this can cause the skill to capture generic requests outside its intended scope and force extra actions or workflows the user did not explicitly request, increasing the risk of prompt/skill overreach and unintended tool use.

Natural-Language Policy Violations

Medium

• Confidence: 92% confidence
• Finding: The skill imposes a mandatory prose style for all user-facing writing, including page copy, without stating that user preferences override it. This can conflict with explicit user instructions, brand requirements, accessibility needs, or localization constraints, causing the agent to disregard higher-priority intent and reducing reliability of the skill's behavior.

Natural-Language Policy Violations

Medium

• Confidence: 95% confidence
• Finding: The doctrine hard-codes a specific writing style for all user-facing text, including bans on certain phrasing and a required tone, without indicating that user instructions can override it. In an agent skill, this can cause instruction interference: the model may ignore a user's requested audience, accessibility, locale, brand voice, or compliance-required wording in favor of the doctrine's preferred style.

Intent-Code Divergence

Medium

• Confidence: 97% confidence
• Finding: This is an explicit contradiction between the component's documented intent and its actual behavior as described in the writeup. The comments claim interactive mouse behavior, but the implementation shown at L52-L53 resets fixed values each frame and never reads pointer input.

Behavior Manipulation

Medium

• Category: Prompt Injection

Content

Gradient text, which impeccable's craft floor bans outright (`skill-ban-gradient-text`, emphasis
comes from weight or size). The component's `children` slot is the escape hatch and the library
should always use it rather than the `title1` / `title2` props.

Conclusion: the arsenal is worth having, and every component in it needs a taming pass before it
touches a client surface. That pass is a skill.

• Confidence: 70% confidence
• Finding: Subtle instructions detected that may alter agent decision-making or introduce hidden biases.

Natural-Language Policy Violations

Medium

• Confidence: 88% confidence
• Finding: The file presents German copy as the primary 'Sounds right' register and only later provides partial English equivalents, which can be read as prescribing a specific language style. While the dealership context is European, the file does not explicitly state that German is required or that language should match the user's locale, so this creates a locale-policy concern.

Natural-Language Policy Violations

Low

• Confidence: 79% confidence
• Finding: The natural-language comment and type definition constrain supported commerce locales to EUR, USD, GBP, and CHF and state that EU stores are inclusive while US stores are not. This can violate locale/language policy expectations if the skill is applied more broadly, because it bakes in regional assumptions without documenting opt-in or scope limitations.

Static analysis

No suspicious patterns detected.