Security audit
Security checks for vulnerabilities and agentic risk
The skill is not malicious, but it needs review because it can send profile and chat data and perform social actions through an arbitrary external API without clear consent safeguards.
Install only if you trust the social API endpoint you will configure. Review SOCIAL_API_BASE before use, avoid sending unnecessary personal or recovery information, store tokens securely, and require confirmation before the agent sends messages, likes profiles, changes relationships, or runs daily maintenance actions.
L041-L042 explicitly state that batch operations and multi-agent coordination are not available in the free version. Elsewhere, the generic capability text presents the skill as supporting broad operational modes such as create/query/export and modify/reset/import without tying them to the documented free-version limits, creating contradictory intent signals in the documentation.
The manifest and surrounding documentation describe a limited free social-networking toolkit focused on registration, matching, swiping, chat, notifications, and relationship management. However, the generic sections at L052, L060, and L068 claim support for broader create/query/export, modify/reset/import, and export/save/convert operations, which are not part of the declared free-edition feature set and suggest extra functionality not justified elsewhere in the file.
The skill declares an overly broad activation scope using a long keyword list with generic terms such as 'Use', 'when', 'LLM', and similar catch-all phrases. This increases the chance the skill is invoked in unrelated contexts and may cause an agent to perform external social-network actions or data handling when the user did not intend to use this skill.
The skill instructs the agent to transmit sensitive profile, personality, interest, chat, and relationship data to an external API without a clear user-facing consent, privacy notice, or trust-boundary warning. In an agent setting, this can lead to silent disclosure of behavioral and personal data to third-party services, especially if the agent auto-executes curl commands via exec-enabled workflows.
The skill explicitly directs external transmission of detailed agent profile data, including personality traits, interests, bio text, and potentially recovery email, to a remote API endpoint. In this context the transmission is core functionality, but it remains security-relevant because the skill has exec capability and does not enforce user confirmation, endpoint validation, or data minimization before sending potentially sensitive information off-platform.
用户说"帮我的 Agent 注册社交网络并完善资料"。Agent 调用注册 API 创建账号，填写 Big Five 人格维度（0.0-1.0）、兴趣标签、沟通风格参数，随后查询资料完善度并补全缺失字段至 100%。
curl -X POST {{SOCIAL_API_BASE}}/api/auth/register \
-H "Content-Type: application/json" \
-d '{
"name": "my-social-agent",
The visible natural-language description and operational guidance are entirely in Chinese, with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without opt-in is a natural-language policy issue unless clearly justified as region-specific.