Security audit
Security checks across malware telemetry and agentic risk
This thesis-writing skill is mostly coherent, but it contains unsafe local execution paths and under-scoped persistent automation that warrant Review before installation.
Install only if you are comfortable reviewing and fixing the unsafe parsing and path handling first. At minimum, replace eval() with strict JSON parsing, validate paper_name so it cannot escape the workspace, remove or gate the phase1_3_skip path, narrow auto-trigger rules, and avoid storing email or confidential thesis material unless you understand where local files and LLM calls will go.
# 尝试提取 JSON（可能在 markdown 代码块中）
json_match = re.search(r'\{[\s\S]*\}', response)
if json_match:
result = eval(json_match.group()) # 安全的简单解析
else:
result = eval(response) # 直接解析
except Exception:if json_match:
result = eval(json_match.group()) # 安全的简单解析
else:
result = eval(response) # 直接解析
except Exception:
return {
"ok": False,try:
json_match = re.search(r'\{[\s\S]*\}', response)
if json_match:
result = eval(json_match.group())
else:
result = eval(response)
return {"ok": True, **result}if json_match:
result = eval(json_match.group())
else:
result = eval(response)
return {"ok": True, **result}
except Exception as e:
return {if json_match:
result = eval(json_match.group()) # 安全的简单解析
else:
result = eval(response) # 直接解析
except Exception:
return {
"ok": False,if json_match:
result = eval(json_match.group())
else:
result = eval(response)
return {"ok": True, **result}
except Exception as e:
return {read -p "作者姓名: " AUTHOR_NAME
# 写入 config.env
cat > "$SKILL_DIR/config.env" << EOF
# MBA Thesis Workflow 配置（自动生成）
WORKSPACE_ROOT=$WORKSPACE_ROOTread -p "作者姓名: " AUTHOR_NAME
# 写入 config.env
cat > "$SKILL_DIR/config.env" << EOF
# MBA Thesis Workflow 配置（自动生成）
WORKSPACE_ROOT=$WORKSPACE_ROOT
USER_EMAIL=$USER_EMAIL63/63 vendors flagged this skill as clean.
Detected: suspicious.dynamic_code_execution