← 返回 Skills 市场

Main Image Editor

Name: Main Image Editor
Author: dhrxy

作者 dhrxy · GitHub ↗ · v1.0.1 · MIT-0

cross-platform ⚠ suspicious

373

总下载

当前安装

版本数

在 OpenClaw 中安装

/install main-image-editor

功能描述

Orchestrate screenshot + Chinese instruction into PSD batch edits with transaction rollback by reusing psd-automator.

使用说明 (SKILL.md)

main-image-editor

Orchestration layer for "main image modification" workflows:

Parse screenshot + Chinese instruction into structured edit tasks.
Convert tasks into psd-automator task JSON.
Execute serially with transaction rollback (rollback all on any failure).
Reuse psd-automator export + bundle output.

Usage

node skills/main-image-editor/scripts/run-main-image-editor.js \
  --request skills/main-image-editor/examples/request.sample.json \
  --index ~/.openclaw/psd-index.json

Dry-run or force execute:

node skills/main-image-editor/scripts/run-main-image-editor.js \
  --request skills/main-image-editor/examples/request.sample.json \
  --dry-run

node skills/main-image-editor/scripts/run-main-image-editor.js \
  --request skills/main-image-editor/examples/request.sample.json \
  --force

Request payload

request JSON supports:

text: Chinese requirement text
screenshotPath: screenshot path (used for match + audit)
confidenceThreshold: optional (default 0.8)
tasks: optional pre-parsed tasks (skip regex parsing when provided)
execution: optional runtime controls (indexPath, dryRun, force)

When fileHint is missing, parser will try to extract .psd/.psb filename from:

request text
OCR text from screenshotPath (local tesseract command or macOS Vision OCR fallback)

Failure policy

Default policy is rollback_all.
Any single PSD failure restores all touched PSD files from transaction backups.
Temporary outputs generated in the failed run are cleaned up before returning.

安全使用建议

This skill appears to implement a PSD orchestration tool, but review these points before installing: - Dependencies: The code calls tesseract (OCR) and, on macOS, runs a Swift Vision snippet; ensure you have tesseract and/or a macOS environment with swift if you expect OCR to work. The metadata does not declare these binaries. - Missing package: The code imports '../../psd-automator' modules that are not included in the bundle. Verify that psd-automator is present and trustworthy in your environment before running. - Filesystem effects: The skill will read arbitrary screenshot/PSD paths, copy PSD files to temp, create backups named with '.txn-bak.TIMESTAMP', and may overwrite originals (unless dry-run). Only run it on files you trust or inside a sandbox. - Risk mitigation: Test with dry-run and non-critical PSDs first, review the psd-automator code that will be executed, run in a restricted environment (container or VM) if possible, and confirm tesseract/swift binaries are from trusted sources. - No network exfiltration was observed in the skill itself, but the psd-automator dependency could perform other actions — inspect that dependency before trusting this skill.

功能分析

Type: OpenClaw Skill Name: main-image-editor Version: 1.0.1 The skill bundle is an orchestration layer for automating Photoshop (PSD/PSB) file edits based on Chinese text instructions and screenshots. It uses standard Node.js APIs and external utilities (Tesseract OCR and a Swift-based macOS Vision OCR script) to parse user intent and execute tasks via a sibling tool, `psd-automator`. The code includes sophisticated logic for transactional safety, including file backups and rollbacks in `lib/task-orchestrator.js`. While it performs high-risk operations such as process execution (`spawnSync`) and file system manipulation, these are necessary for its stated functionality, and there is no evidence of malicious intent, data exfiltration, or unauthorized persistence.

能力评估

⚠ Purpose & Capability

The code implements PSD editing orchestration that matches the description (parsing Chinese instructions, OCR, converting to psd-automator tasks, transactional rollback). However metadata claims no required binaries or dependencies while the code calls external binaries (tesseract, swift) and imports local modules from a separate '../../psd-automator' package that is not included in the skill bundle or install spec. The omission of those requirements is an incoherence the user should be aware of.

ℹ Instruction Scope

SKILL.md instructs running the provided node script with a request JSON and index; the runtime instructions and code align. The code will read screenshots and PSD/PSB files (including expanding ~ to the user's home), run OCR on local files, copy PSDs to local temp, create backups ('.txn-bak.TIMESTAMP'), and overwrite originals when not in dry-run. Those file reads/writes are expected for the stated purpose but grant the skill broad local filesystem access to any paths supplied in requests or discovered via OCR.

ℹ Install Mechanism

There is no install spec (lowest install risk), and all code is included. That keeps risk moderate, but the skill expects a separate 'psd-automator' module reachable via relative path and relies on host binaries (node, tesseract, and optionally swift on macOS). These runtime dependencies are not declared in the registry metadata or SKILL manifest.

ℹ Credentials

No environment variables or credentials are requested (good). But the code spawns child processes for 'tesseract' and 'swift' and may call 'node' to run the psd-automator runner — these runtime binaries are required by the implementation even though 'required binaries' is empty in the metadata. Also the skill will read and write files under the user's home and other arbitrary paths referenced in requests or OCR.

✓ Persistence & Privilege

The skill does not request persistent inclusion (always: false) and does not modify other skills or global agent settings. It creates temporary files under OS temp and writes backups next to PSD files; that is expected for transactional editing and is limited to file-level effects rather than agent-level persistence.

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 main-image-editor
安装完成后，直接呼叫该 Skill 的名称或使用 /main-image-editor 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

v1.0.1

maintenance update

v1.0.0

Initial release: PSD main image orchestration with screenshot parsing and transaction rollback

元数据

Slug main-image-editor

版本 1.0.1

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 2

常见问题