360 Secure SkillHub
360 安全技能中心

Psd Batch Export

核心功能含PSD文本层分析、Excel/CSV数据与图层映射、效果预览、批量导出可编辑PSD及高清PNG,支持结果校验、字体/OCR依赖检查,内置门票、证书等模板。适用于批量制作营销、办公物料。通过CLI操作,输入PSD模板/内置ID、数据文件,输出成品文件及结构化报告,可提前诊断环境依赖。

luis1213899
luis1213899
热度 57.79
最近更新 2026-06-06 13:51:38
分类:效率依赖:代码执行 / 读写文件应用场景:batch image generation / 数据处理 / 自动化脚本 / 模板 ...

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add psd-batch-export
Metadata
name
psd-batch-export
description
Production PSD batch workflow for analyzing PSD text layers, mapping Excel/CSV data to editable layers, previewing, exporting editable PSD plus PNG batches, validating results, checking fonts/OCR dependencies, and using built-in PSD templates for tickets, certificates, invitations, badges, posters, daily cards, and recruitment assets. Use whenever the user mentions PSD, Photoshop templates, layer text replacement, batch image generation, Excel-driven design output, PSD analysis, print-ready PNG export, or OpenClaw/Claude/Codex skill packaging for PSD automation.
allowed-tools
["Read","Write","Edit","Bash"]

PSD Batch Export Skill

Use this skill for production PSD batch generation. Resolve all paths relative to the directory containing this SKILL.md; do not rely on machine-specific absolute paths.

Preferred Entry Point

Use the unified CLI first:

python scripts/psd_batch.py diagnose --json
python scripts/psd_batch.py analyze "template.psd" --data "data.xlsx" --json-out report.json
python scripts/psd_batch.py preview "template.psd" --data "data.xlsx" --out "out" --rows 3
python scripts/psd_batch.py export "template.psd" --data "data.xlsx" --out "out" --dpi 300 --verify-samples 3
python scripts/psd_batch.py verify "out" --samples 3
python scripts/psd_batch.py templates list
python scripts/psd_batch.py templates info morning

Built-in template IDs can be used instead of PSD paths: morning, tech, doodle.

Workflow

  1. Run diagnose --json before production work to check required Python packages, fonts, templates, optional OCR, and optional LLM configuration.
  2. Run analyze with the PSD and optional Excel/CSV data source. Review layers, mapping, capacity_risks, fonts, warnings, and errors.
  3. Run preview before full export. Check out/dryrun_preview/ and the JSON report.
  4. Run export to generate out/psd/, out/png/, and out/verify_report/report.json. Replacement text keeps the original PSD text-layer color; pass --color R G B only when the user explicitly requests a uniform override.
  5. Run verify again when outputs are moved or inspected separately.

Use --strict for production batches where low-confidence column matches, capacity overflow, or missing fonts should stop the run before export.

Stable Report Contract

All primary commands support --json and --json-out <path>. The report includes:

  • status: ok or error
  • inputs: PSD, data, output, or template inputs
  • layers: detected PSD text layer details
  • mapping: data-column to layer mapping with confidence
  • capacity_risks: text fields likely to truncate
  • fonts: available fonts and matching recommendations
  • outputs: generated directories, files, and verification report paths
  • warnings: non-fatal production risks
  • errors: fatal failures
  • runtime_sec: elapsed seconds

Exit codes are stable: 0 success, 1 input error, 2 dependency/PSD compatibility error, 3 export or verification failure, 4 unexpected error.

Legacy Entrypoints

Older scripts are kept for compatibility:

python scripts/batch_from_excel.py "data.xlsx" "template.psd" "out" --dry-run
python scripts/batch_from_excel.py "data.xlsx" "template.psd" "out"
python scripts/ai_designer.py --list
python scripts/color_tools.py --recommend ticket
python scripts/smart_layout.py --preset certificate
python scripts/psd_text_editor.py "in.psd" "out.psd" "图层名=新文本"

Prefer scripts/psd_batch.py for new workflows because it provides structured reports and stable exit codes.

References

  • references/templates.json: built-in PSD template catalog and editable fields.
  • references/runtime-compat.md: Claude, Codex, OpenClaw, and ClawHub install/validation commands.

Verification

Run these before publishing or sharing the skill:

python -m compileall -q scripts
python -m pytest -q
python scripts/psd_batch.py diagnose --json

For Codex validation on Windows, run with UTF-8 mode:

$env:PYTHONUTF8='1'; python C:\Users\26240\.codex\skills\.system\skill-creator\scripts\quick_validate.py .

For OpenClaw validation, install the local skill and then check visibility:

openclaw skills install . --global --force
openclaw skills check --json

Do not publish to ClawHub unless the user explicitly asks for a release.