← Back to Skills Marketplace
chenqu108

LC Scene Analysis HTTP

by chenqu108 · GitHub ↗ · v1.0.5 · MIT-0
windowslinux ✓ Security Clean
193
Downloads
0
Stars
0
Active Installs
6
Versions
Install in OpenClaw
/install lc-scene-analysis-http
Description
工地隐患分析模型技能,默认处理施工现场静态图片的隐患识别;仅在用户明确要求时调用数据采集或交叉验证接口。
README (SKILL.md)

\r \r

LC Scene Analysis HTTP\r

\r 这是一个工地隐患分析模型技能,用于处理施工现场静态图片的安全隐患识别任务。\r \r

何时使用\r

\r 当用户提供施工现场图片或图片 URL,并希望:\r \r

  • 识别图片中的安全隐患\r
  • 获取模型分析结果\r
  • 输出隐患结论或简要说明\r \r 此时使用本技能。\r \r

不要使用\r

\r 以下情况不要调用本技能:\r \r

  • 视频流分析\r
  • 摄像头实时拉流\r
  • 本地模型推理\r
  • PLC、机器人、设备控制\r
  • 非施工现场图片分析\r
  • 与工地隐患识别无关的任务\r \r

前置要求\r

\r 需要存在 auth profile:\r \r

  • lc_scene_http:default\r \r 该 profile 需提供以下字段:\r \r
  • api_base:服务基地址\r
  • api_key:接口认证信息\r
  • flow_id:默认 flowId\r
  • algorithm_id:默认 algorithmId,可为空\r \r 如果用户本次请求显式提供了 flow_idalgorithm_id,优先使用用户输入;否则使用默认配置。\r \r 不要向用户展示任何密钥、token、profile 内容或本地认证文件内容。\r \r

接口说明\r

\r

1. 默认接口:智能巡检\r

接口:\r \r POST /api/chat/agent/chatOnceNew\r \r 用途:\r \r

  • 默认图片隐患分析\r
  • 工地安全隐患识别\r
  • 普通分析请求优先走该接口\r \r

2. 数据采集接口\r

接口:\r \r POST /api/chat/agent/chatOnceRaw\r \r 用途:\r \r

  • 用户明确要求数据采集\r
  • 用户明确要求结构化结果、原始结果、JSON 结果时调用\r \r

3. 交叉验证接口\r

接口:\r \r POST /api/chat/agent/crossVerify\r \r 用途:\r \r

  • 用户明确要求交叉验证、复核、二次校验时调用\r
  • 当前会结合 CV 小模型对输出结果进行二次校验\r \r

选择规则\r

\r 按以下规则选择接口:\r \r

  • 用户明确要求“交叉验证 / 复核 / 二次校验 / 再确认” → crossVerify\r
  • 用户明确要求“数据采集 / 原始结果 / 结构化结果 / JSON 结果” → chatOnceRaw\r
  • 其他普通工地隐患分析请求 → chatOnceNew\r \r 如果无法判断,默认使用 chatOnceNew。\r \r

输入要求\r

\r 优先接受以下输入:\r \r

  • 单张图片 URL\r
  • 多张图片 URL\r
  • 用户额外提供的 flow_id\r
  • 用户额外提供的 algorithm_id\r \r 如果没有图片或图片 URL,不要调用接口,应明确告知缺少输入。\r \r

请求要求\r

\r 调用时:\r \r

  • lc_scene_http:default 读取默认配置\r
  • 使用其中的 api_baseapi_keyflow_idalgorithm_id\r
  • 若用户显式传入 flow_id / algorithm_id,则覆盖默认值\r
  • 默认调用 chatOnceNew\r
  • 仅在用户明确要求时调用 chatOnceRawcrossVerify\r \r

输出要求\r

\r

  • 默认返回核心隐患识别结果\r
  • 如用户要求原始结果或结构化结果,尽量保留关键字段\r
  • 可以做简洁总结,但不要篡改关键结论\r
  • 不确定内容不要说成确定事实\r \r

错误处理\r

\r 以下情况要明确报错,不要伪造结果:\r \r

  • auth profile 不存在\r
  • profile 缺少必要字段\r
  • 图片输入缺失\r
  • 服务返回 4xx / 5xx\r
  • 服务超时\r
  • 返回结构异常\r \r

安全要求\r

\r

  • 不要输出密钥、token、Authorization 信息\r
  • 不要输出认证文件内容\r
  • 对外部图片 URL 保持谨慎\r
  • 如果任务明显超出工地隐患分析范围,不要强行调用本技能\r \r

示例\r

\r

示例 1:默认分析\r

用户:\r “帮我分析这张施工现场图片有哪些隐患:https://example.com/a.jpg”\r \r 处理:\r

  • 调用 chatOnceNew\r \r

示例 2:数据采集\r

用户:\r “把这张图按结构化结果返回:https://example.com/b.jpg”\r \r 处理:\r

  • 调用 chatOnceRaw\r \r

示例 3:交叉验证\r

用户:\r “这张图再帮我做一次交叉验证:https://example.com/c.jpg”\r \r 处理:\r

  • 调用 crossVerify
Usage Guidance
This skill appears to do what it says: call a remote HTTP API to analyze construction-site images. Before installing, verify the auth profile and api_base are correct and trusted (ensure api_base points to a domain you control or a reputable provider). Restrict the api_key to minimal scope and treat it as a secret (store it in the platform's secure config). Ask the publisher to fix the registry metadata (the 'Required config paths' blob and version mismatch) so it's explicit that an auth profile with api_key is required. Consider testing with non-sensitive images first and confirm the provider's privacy policy for image storage/retention. If you cannot confirm api_base ownership or the publisher's identity, treat the skill as higher risk and avoid providing sensitive images or credentials.
Capability Analysis
Type: OpenClaw Skill Name: lc-scene-analysis-http Version: 1.0.5 The lc-scene-analysis-http skill is designed to analyze construction site images for safety hazards by interacting with an external API. It uses curl to communicate with endpoints like /api/chat/agent/chatOnceNew and includes explicit instructions in SKILL.md to prevent the leakage of API keys or sensitive configuration data. The logic is consistent with its stated purpose and lacks any indicators of malicious intent or unauthorized data access.
Capability Assessment
Purpose & Capability
Name/description (工地隐患分析模型) match the runtime instructions: it calls three HTTP endpoints to perform analysis, data-collection, and cross-validation. Requiring curl and an auth profile with api_base/api_key is appropriate for this purpose.
Instruction Scope
SKILL.md confines behavior to static image analysis, specifies when to call each endpoint, and explicitly forbids unrelated actions (video streams, device control, local model inference). It also instructs not to reveal keys/tokens. There are no instructions to read unrelated system files or exfiltrate arbitrary data.
Install Mechanism
Instruction-only skill with no install spec and no code files — lowest install risk. It only requires curl to be present, which is reasonable for issuing HTTP requests.
Credentials
The skill requires an auth profile containing api_base and api_key (and flow_id/algorithm_id). That credential request is proportionate to calling a remote analysis API. However, the registry metadata is inconsistent/opaque: top-level shows 'Required config paths: [object Object]' and lists 'Required env vars: none' / 'Primary credential: none', which fails to clearly declare the needed profile/credential. This metadata mismatch reduces transparency and should be corrected.
Persistence & Privilege
always=false, user-invocable, and no install-time persistence. The skill does not request elevated/system-wide privileges or modify other skills' configs.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install lc-scene-analysis-http
  3. After installation, invoke the skill by name or use /lc-scene-analysis-http
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.5
- 增加了对 Windows 操作系统的显式支持说明 - 在 metadata 中添加了 auth profile 配置需求说明 - 其他内容保持不变
v1.0.4
- No user-facing or functional changes in this version. - No file or documentation updates detected.
v1.0.3
No user-facing changes detected in this version.
v1.0.2
No user-facing changes detected in this version. - No file or documentation changes between versions 1.0.0 and 1.0.2.
v1.0.1
- Simplified configuration: replaced previous secrets file method with an auth profile system (`lc_scene_http:default`). - Only the README.md file added; removed _meta.json and secrets.json. - New error handling guidance: explicit reporting for missing profiles, inputs, or API errors. - SKILL.md updated with clearer usage boundaries, input requirements, and concise interface selection rules. - Security rules tightened—API keys and sensitive info will never be shown to users. - Use case examples and default behaviors are now easy to follow.
v1.0.0
- 首发版本,支持施工现场图片的安全分析。 - 通过远程 HTTP 服务实现智能巡检、数据采集、交叉验证三大功能,适用多类业务场景。 - 自动从 OpenClaw 配置文件读取必要 API 参数,支持用户自定义覆盖。 - 根据用户需求智能选择检测、采集或复核接口。 - 提示并阻止操作当配置缺失,避免密钥明文泄露。 - 支持用 curl 与 powershell 实现接口调用,适配主流系统。
Metadata
Slug lc-scene-analysis-http
Version 1.0.5
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 6
Frequently Asked Questions

What is LC Scene Analysis HTTP?

工地隐患分析模型技能,默认处理施工现场静态图片的隐患识别;仅在用户明确要求时调用数据采集或交叉验证接口。 It is an AI Agent Skill for Claude Code / OpenClaw, with 193 downloads so far.

How do I install LC Scene Analysis HTTP?

Run "/install lc-scene-analysis-http" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is LC Scene Analysis HTTP free?

Yes, LC Scene Analysis HTTP is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does LC Scene Analysis HTTP support?

LC Scene Analysis HTTP is cross-platform and runs anywhere OpenClaw / Claude Code is available (windows, linux).

Who created LC Scene Analysis HTTP?

It is built and maintained by chenqu108 (@chenqu108); the current version is v1.0.5.

💬 Comments