Security audit

Cocktail Craft

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a cocktail reference skill with no evidence of hidden code execution, credential access, exfiltration, or destructive behavior, though users should review installation commands and food/alcohol safety gaps.

Install only after reviewing the GitHub source and the exact commands being run, especially because they write persistent skill files locally. When using the skill, treat homemade infusions, potent botanicals, high-proof spirits, and raw egg recipes as needing separate food-safety and responsible-adult-use checks.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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 (7)

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: The README explicitly instructs an agent to download code from GitHub and move files into local skill directories, which introduces network-fetching and filesystem modification behavior unrelated to the cocktail-reference content itself. In an agent context, these commands can be executed with user trust and may install unreviewed or mutable remote content, creating supply-chain and local-environment risks.

Vague Triggers

Medium

Confidence: 80% confidence
Finding: The trigger description says the skill should activate whenever users discuss cocktail recipes, techniques, or liquor history, which is broad enough to match many ordinary conversations. Overbroad auto-triggering can cause unnecessary invocation of the skill, increasing the chance an agent follows its instructions or treats its content as authoritative in contexts where the user did not intend that behavior.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The installation section provides shell commands that download, unzip, clone, and move files into agent directories without warning that they change the local environment and rely on remote content. In an agent setting, omission of these warnings increases the risk that users or agents execute the commands without understanding trust, persistence, and overwrite implications.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The skill explicitly states that any cocktail-related question must first load the entire references directory before answering, and the trigger list spans nearly all cocktail use cases. This creates an unnecessarily broad activation surface and can cause excessive context loading, routing hijack of general food-and-drink queries, and degraded system behavior even when a narrower response would suffice.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The file gives actionable instructions for preparing high-proof alcohol infusions and extracts, including >50% ABV and specific products, but omits basic safety guidance on flammability, ventilation, labeling, storage, and consumption risks. In a consumer-facing cocktail skill, this can lead users to unsafe handling of high-proof spirits or inappropriate ingestion of homemade extracts.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The document recommends botanicals such as cinchona bark and wormwood and notes quinine in cinchona, but provides no warning about dosage, contraindications, sourcing quality, or regulatory/consumption considerations. Because this is a homemade-ingredients guide for direct beverage use, users may assume all listed botanicals are safe to source and consume without limits.

Natural-Language Policy Violations

Low

Confidence: 90% confidence
Finding: The CSV includes cocktail recipes using raw egg ingredients such as egg white and egg yolk without any accompanying safety warning, handling guidance, or user opt-in. In a consumer-facing cocktail skill, this can expose users to foodborne illness risk, especially for immunocompromised users, pregnant users, children, or users unaware of the hazard.

Static analysis

No suspicious patterns detected.