Security audit

fasaha

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Arabic proofreading skill with broad activation and small local self-updating notes, but no hidden, destructive, credential, or exfiltration behavior was found.

Installers should know this skill may activate automatically for longer Arabic output and may update its own local glossary, register profile, and correction log. Review those files if you want tighter control over tone preferences or learned terminology.

SkillSpector

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 (2)

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger conditions are very broad and include common Arabic-writing and translation requests, which can cause the skill to activate in many ordinary contexts without explicit user intent. In an agent system, over-broad automatic invocation can override task prioritization, create unnecessary transformations of user content, and increase the chance of the skill being applied where it is not appropriate, especially when it also references self-maintained state files.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The proactive guidance to run the skill before any Arabic prose longer than 2–3 sentences is ambiguous and expansive, making the skill effectively self-invoking across a large class of normal outputs. This can lead to unintended rewriting of model responses, conflict with other skills or user instructions, and hidden modification of tone/register beyond what the user asked for.

Static analysis

No suspicious patterns detected.