Flame & Smoke Detection Skill | 烟火检测技能

Detects fire and smoke in video scenes. Supports both video stream and image analysis. Suitable for fire early warning scenarios such as security surveillanc...

What to check before installing or running this skill: - Trust the API endpoints: open skills/smyx_common/scripts/config.yaml and config-dev/test/prod to confirm base URLs and any api-key values. The config references lifeemergence.com and dev/test hosts — only use the skill if those services are trusted. - Inspect network utility code: review skills/smyx_common/scripts/util.py (RequestUtil) to see exactly what headers, authentication, and destinations are used for HTTP requests and whether any unexpected metadata or environment values are sent. - Confirm data handling and retention: the package will read local media files, may save attachments into the skill workspace, and smyx_common/dao.py creates a local SQLite DB under the workspace/data directory. If you will upload camera footage, ensure storage locations and retention are acceptable for privacy/security. - Credentials and config gap: although metadata lists no required env vars, the code reads environment variables and YAML config for API keys and open-id. Ensure you supply only intended credentials and that they are stored securely (or run without persisted credentials if possible). - Unrelated code included: the repository includes face-analysis and pet/health references and large common libraries. That suggests code reuse and increases attack surface; review whether any of those components expose extra endpoints or behaviors you don't want. - Sandbox first: run the skill in an isolated environment (no access to sensitive networks or production data) and observe outbound network calls before giving it access to real video or credentials. If you need, I can: (1) list the exact config file paths and values found in the repo, (2) extract and summarize RequestUtil behavior to show what outbound data is sent, or (3) point to the lines that create/modify the local DB and where attachments are saved.

Type: OpenClaw Skill Name: smyx-fire-smoke-detection-analysis Version: 1.0.0 The skill bundle uses aggressive prompt injection techniques in SKILL.md (e.g., 'Mandatory Memory Rules', 'Highest Priority') to force the AI agent to ignore its internal memory and exclusively use a specific external API (lifeemergence.com). The underlying framework in smyx_common/scripts/util.py and dao.py automatically performs a remote 'phoneLogin' using the user-provided open-id and persists session tokens in a local SQLite database. Furthermore, skill.py contains logic to execute shell commands via subprocess.run to invoke the 'openclaw' agent recursively, which is a high-risk capability for an agent skill.