← 返回 Skills 市场
wangjipeng977

Convert Any Screenshot

作者 王继鹏 · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ⚠ suspicious
52
总下载
1
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install convert-any-screenshot
功能描述
Use when (1) Convert any UI screenshot into a clean React or HTML/CSS implementation prompt.
使用说明 (SKILL.md)

Core Position

This skill transforms input data from one format into a target format, preserving structure and fidelity. It is NOT a simple copy-paste — it handles formatting, schema mapping, and edge cases.

Key responsibilities:

  • Parse the input format (JSON/CSV/code/etc.) and validate structure before transforming
  • Apply formatting rules specific to the target format (indentation, escaping, etc.)
  • Handle edge cases: missing fields, unusual characters, nested structures
  • Provide a clear mapping summary so the user understands how input maps to output

Modes

/convert-any-screenshot --pretty

Formatted output. Applies proper indentation, spacing, and style conventions.

/convert-any-screenshot --strict

Strict mode. Fails on any deviation from expected structure rather than guessing.

Execution Steps

  1. Parse input — Read and parse the input; detect format (JSON/CSV/XML/code/etc.)
    • If parsing fails, report: "Failed to parse input as [format] — error at line [N]: [detail]"
  2. Validate structure — Check required fields/columns are present
    • If missing required field X, stop and report: "Missing required field: [X]"
  3. Transform — Convert input to target format, applying format-specific rules
    • Preserve all data — do not silently drop fields
    • Apply proper escaping for special characters (quotes, newlines, etc.)
  4. Validate output — Run the target format parser on the result to confirm it's valid
    • If output is invalid, revert to previous version and report what went wrong
  5. Deliver — Return the converted output with a brief mapping summary

Mandatory Rules

Do not

  • Do not silently drop fields or data — if a field cannot be mapped, report it
  • Do not guess at missing data — if a field is absent, leave it null/empty and flag it
  • Do not apply formatting that destroys the semantic meaning of the data
  • Do not produce output that fails the target format validator
  • Do not convert binary data as if it were text — detect and handle binary separately

Do

  • Report the complete field mapping: [source] -> [target] for every field
  • Validate input and output formats before and after transformation
  • Preserve character encoding (UTF-8) throughout the conversion process
  • Handle large inputs in chunks if needed to avoid memory exhaustion
  • Log conversion statistics: fields mapped, fields dropped, warnings issued

Quality Bar

Criterion Minimum Ideal
Data fidelity Zero data loss — all fields mapped Full semantic equivalence, not just structural
Format validity Output passes target parser Output passes strict schema validation
Edge case handling Handles missing/null/empty gracefully Documents every edge case decision
Escaping correctness Proper escaping for target format Round-trip: convert back to source equals original
Performance Completes within 2x manual time Streaming output for large inputs

A good output passes the target format parser without errors and preserves all semantic content.

Good vs. Bad Examples

Scenario Bad Good
Missing field Omits field from output silently Reports "Field [X] absent — output null, flagged as warning"
Special characters Only escapes visible chars Escapes all special chars per target format spec
Large input Loads entire file into memory Streams in chunks, reports progress at 25/50/75%
Output validation Skips validation Runs target parser on output, confirms valid before returning
Format error Returns raw output with error text appended Returns nothing, reports "Output invalid: [parser error] at [location]"
安全使用建议
Review carefully before installing. The files do not show exfiltration, destructive actions, or persistence, but the skill should be revised so its instructions, README, credential claims, and modes all match the screenshot-to-code purpose and clearly state whether any local files can be modified.
能力标签
requires-sensitive-credentials
能力评估
Purpose & Capability
Manifest and README advertise converting UI screenshots into React or HTML/CSS prompts, while SKILL.md primarily instructs a generic JSON/CSV/XML/code transformation workflow; this is a concrete purpose mismatch that could route unrelated or sensitive data into the skill.
Instruction Scope
The activation and execution instructions are broad and contradictory, covering generic parsing, schema mapping, binary detection, chunking, and conversion statistics rather than the screenshot-to-code use case.
Install Mechanism
Installation is a normal ClawHub/manual skill copy flow, and the included setup script only prints usage; however README examples mention a write mode without defining overwrite behavior.
Credentials
The included helper script uses only Python stdlib, reads stdin or a named file, and prints a generated UI implementation prompt; static review found no network calls, package installs, destructive commands, persistence setup, or credential-store access.
Persistence & Privilege
No background worker, daemon, autostart, privilege escalation, or persistent local indexing was found, but README/metadata claim API-key or sensitive-credential relevance without a clear need in the artifacts.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install convert-any-screenshot
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /convert-any-screenshot 触发
  4. 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
v1.0.1
Initial release
元数据
Slug convert-any-screenshot
版本 1.0.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Convert Any Screenshot 是什么?

Use when (1) Convert any UI screenshot into a clean React or HTML/CSS implementation prompt. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 52 次。

如何安装 Convert Any Screenshot?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install convert-any-screenshot」即可一键安装,无需额外配置。

Convert Any Screenshot 是免费的吗?

是的,Convert Any Screenshot 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Convert Any Screenshot 支持哪些平台?

Convert Any Screenshot 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 Convert Any Screenshot?

由 王继鹏(@wangjipeng977)开发并维护,当前版本 v1.0.1。

💬 留言讨论