← 返回 Skills 市场
Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能
作者
smyx-sunjinhui
· GitHub ↗
· v1.0.0
· MIT-0
64
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install smyx-smoking-detection-analysis
功能描述
Automatically detects smoking behavior in target areas based on computer vision; supports real-time detection of video streams, images, and video files; iden...
功能分析
Type: OpenClaw Skill
Name: smyx-smoking-detection-analysis
Version: 1.0.0
The skill bundle exhibits several high-risk behaviors and architectural patterns. SKILL.md contains 'Mandatory Rules' designed to override the AI agent's default behavior by explicitly forbidding the use of local memory or LanceDB, which is a form of prompt injection used to control the agent's context. The common utility library (smyx_common/scripts/skill.py) includes an AgentSkill class that uses subprocess.run to execute the 'openclaw' command-line tool, creating a risk for recursive command injection. Furthermore, RequestUtil in util.py automatically handles user registration and token management by sending PII (like phone numbers provided as open-id) to external endpoints (lifeemergence.com) and storing credentials in a local SQLite database. While these features may support the cloud-based smoking detection service, the combination of subverting agent memory and providing shell execution capabilities is highly irregular.
能力标签
能力评估
Purpose & Capability
The name/description (smoking detection via CV) matches the included code (scripts/smoking_detection_analysis.py and related API client code). Declared dependency on smyx_common is reasonable because shared APIs/config/utilities are used. However, the package bundle also contains a large 'face_analysis' subskill and a broad 'smyx_common' library (DB/DAO/config) that are not strictly necessary for a focused smoking-detection helper; their presence increases complexity and persistence surface.
Instruction Scope
SKILL.md mandates obtaining an open-id by reading config files in skills/smyx_common/scripts/config.yaml (and workspace-level config), requires saving uploaded attachments to a local attachments directory, and forbids reading local 'memory' files or LanceDB. The code does read config.yaml, sets ConstantEnum.CURRENT__OPEN_ID from passed args or environment, and will read and write files (including creating a local SQLite DB under workspace/data via the DAO utilities). The instructions force uploading media to a remote API (via RequestUtil/http_post), meaning user media is transmitted off-host. The prohibition on local memory access contrasts with the skill's use of other local config files and local DB utilities—this is inconsistent and should be clarified.
Install Mechanism
No install spec is provided (instruction-only), which is low-risk for automatic code fetching; however the repository contains many Python modules and a large requirements list in smyx_common/requirements.txt. Running the skill will require installing many dependencies (including network and DB libs). The lack of an install step means a user or operator will need to review and install dependencies manually in their environment—this elevates operational risk if done without inspection or sandboxing.
Credentials
The registry metadata shows no required env vars, but the code reads several environment variables implicitly: OPENCLAW_SENDER_OPEN_ID / OPENCLAW_SENDER_USERNAME / FEISHU_OPEN_ID (via ConstantEnum.init), and OPENCLAW_WORKSPACE is used to determine where DB/files are stored. The skill also expects/prohibits particular open-id values and requires an open-id to operate. The implicit use of workspace and sender env vars is not declared in the skill metadata, which is a proportionality / transparency problem. The skill will send uploaded media and request parameters to external API endpoints (configured in smyx_common config.yaml), so API keys, personal identifiers (open-id/username/phone), and media may be transmitted off-host.
Persistence & Privilege
The skill will create/use local persistence: it uses dao.py to initialize or access a SQLite DB under a data directory (derived from OPENCLAW_WORKSPACE or workspace path) and will save uploaded attachments to a skill attachments folder. The SKILL.md explicitly instructs saving attachments locally. The skill does not request 'always: true' and does not try to modify other skills' configurations, but it does read other-skill config files (skills/smyx_common/scripts/config.yaml) and can create local files and DBs—this is a non-negligible level of persistence and should be expected and consented to by the operator.
scan_findings_in_context
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install smyx-smoking-detection-analysis - 安装完成后，直接呼叫该 Skill 的名称或使用
/smyx-smoking-detection-analysis触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
- Initial release of the smoking-detection-analysis skill for automated smoking behavior detection using computer vision.
- Supports real-time detection in video streams, images, and video files, identifying violations and triggering alerts.
- Strictly enforces usage of cloud interfaces for historical report queries—local memory or cache retrieval is prohibited.
- Requires open-id (from config or user) before processing any detection or report query; prohibits using defaults or guesses.
- Provides structured analysis reports and Markdown table output for history, including direct links to report images.
- Enforces dependencies and input/output formats for reliable, secure smoking detection and management workflow.
元数据
常见问题
Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能 是什么？
Automatically detects smoking behavior in target areas based on computer vision; supports real-time detection of video streams, images, and video files; iden... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 64 次。
如何安装 Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install smyx-smoking-detection-analysis」即可一键安装，无需额外配置。
Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能 是免费的吗？
是的，Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能 支持哪些平台？
Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Intelligent Public Smoking Detection Skill | 公共场所吸烟行为智能检测技能？
由 smyx-sunjinhui（@smyx-sunjinhui）开发并维护，当前版本 v1.0.0。
推荐 Skills