Security audit
Security checks for vulnerabilities and agentic risk
This skill is a simple scientific image-prompt generator with minor scope/documentation issues but no evidence of credential access, persistence, exfiltration, or destructive behavior.
This appears safe to install for generating scientific journal-cover and graphical-abstract prompts. Users should keep it scoped to visual prompt generation and should not rely on it for broader academic writing or scientific claims without separate review.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
python scripts/cover_prompter.py
--topic "neuroscience synaptic transmission"
--style artistic
--output prompt.txt
---
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
]
}
return prompt
def print_prompt(self, prompt):
"""Print formatted prompt."""
The skill's trigger text expands beyond journal-cover prompting into broad 'academic writing tasks,' which can cause an agent to invoke this skill for unrelated requests. Over-broad activation increases the chance that prompt-generation logic or packaged scripts are used in contexts they were not designed for, weakening task isolation and making downstream misuse more likely.
The finding appears valid because the audit result itself shows evidence of scope drift: one stress scenario failed both 'stays within declared skill scope' and 'required research safety/boundary guidance.' A broad invocation description can cause the skill to be selected for generic academic-writing tasks, increasing the chance it produces outputs outside its intended journal-cover/scientific-artwork domain and bypasses domain-specific safeguards.
No suspicious patterns detected.