Security audit

Caveman

Security checks for vulnerabilities and agentic risk

Overview

This skill only changes the assistant into a terse response style and does not add code execution, data access, or external communication.

Install this only if you want the assistant to stay in a terse style after broad prompts like "be brief" or "less tokens." Use "normal mode" or "stop caveman" to turn it off, and avoid the wenyan modes unless you specifically want Classical Chinese-style output.

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: 94% confidence
Finding: The trigger phrases are broad enough to activate on common user requests like 'be brief' or 'less tokens', which can silently switch the assistant into a persistent alternate response mode without clear user intent. In a security-sensitive context, this increases the chance that warnings, nuance, or policy-required detail are compressed or omitted, even though the skill later tries to mitigate this with an Auto-Clarity section.

Natural-Language Policy Violations

Medium

Confidence: 87% confidence
Finding: The skill defines Classical Chinese output variants as supported modes and presents them alongside normal intensity levels, creating a realistic path for output language switching without clear, informed user opt-in. That can cause accessibility, comprehension, and locale-policy issues, especially when the skill also supports auto-triggering and persistence, making it easier for the assistant to respond in an unexpected language/register.

Static analysis

No suspicious patterns detected.