clip-organizer - SKILL.md Agent Skill

name: clip-organizer description: Organize dated clip folders created by Web Collector into stable topic folders, generate a local daily digest, and clean up the processed date inbox. Use when Codex needs to整理剪藏文件, 按日期归档收藏, 生成剪藏日报, or 把 web collector 收藏内容分类 for a local Markdown knowledge base.

Clip Organizer

Use this skill to turn 剪藏文件/YYYY-MM-DD/ date folders into a maintained local archive.

The skill is designed for a workflow where Web Collector drops raw Markdown files into date-named inbox folders, and Codex needs to classify them, write a digest, move them into long-term category folders, and delete the processed date folder only after the whole batch succeeds.

Default Workflow

Confirm the workspace root.
Run python3 scripts/organize_clips.py --workspace "<workspace>" --latest for the newest date folder, or replace --latest with --date YYYY-MM-DD / --all-pending.
Start with --dry-run when validating a new batch or after changing rules.
Review the printed plan or generated digest summary.
Re-run without --dry-run to write:
- 05-素材收集/clip-digest-YYYYMMDD.md
- 剪藏文件/.clip-staging/YYYY-MM-DD-manifest.json
Ensure the date folder is removed only after every file in that batch was handled successfully.

What The Scripts Do

scripts/organize_clips.py Main entrypoint. Loads clips from one or more date folders, classifies them, generates the digest, moves files into category folders, and removes the processed date directory on success.
scripts/classify_clip.py Inspect one Markdown clip and print its inferred category, subcategory, scores, and reasons as JSON.
scripts/build_digest.py Render a digest Markdown file from a manifest JSON file.
scripts/common.py Shared parsing, classification, scoring, digest, and path helpers.

Default Workspace Contract

This skill defaults to the current writing workspace:

Workspace: /Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目
Input date folders: 剪藏文件/YYYY-MM-DD/*.md
Digest output: 05-素材收集/clip-digest-YYYYMMDD.md
Manifest cache: 剪藏文件/.clip-staging/YYYY-MM-DD-manifest.json

Pass --workspace explicitly whenever you use the scripts. Do not hardcode assumptions in the calling prompt when a different workspace is intended.

Classification Rules

Read taxonomy.md before changing routing rules.

The scripts keep the existing top-level folders and only use these extra subfolders:

📖 技术教程/知识管理与记忆
🔧 工具推荐/开发环境与编程工具
📌 其他/待复核异常抓取

Classification order is fixed:

Promotion packs, malformed captures, README-like artifacts, or event material
Hands-on case studies, build logs, workflow retrospectives, business cases
Tutorials, explainers, architecture posts, official deep dives
Tools, products, skills, plugins, platforms, resource lists
Industry trend or company/market judgment
General ideas, commentary, and topic sparks

Digest Rules

Read digest-format.md before changing report structure.

The digest currently contains:

## 1. 今日创作建议
### 1.1 必须要学习的
### 1.2 必须去实践的
## 2. 📊 当天收藏总览
## 3. 全量收藏清单

The recommendation sections are creator-oriented. Display fields are intentionally compact and currently prefer frontmatter metadata.

The scripts compute creator-facing scores and labels, including:

learning_score
execution_score
topic_fit_score
creative_priority_score
creative_value_label

Selection behavior:

必须要学习的 prefers clips that are good for concept, framework, and judgment intake
必须去实践的 prefers clips that are good for tool testing, workflow replay, and case-study reproduction
full-list duplicates are collapsed to one main display entry

Commands

python3 scripts/organize_clips.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --latest
python3 scripts/organize_clips.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --date 2026-04-07
python3 scripts/organize_clips.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --all-pending
python3 scripts/organize_clips.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --date 2026-04-07 --dry-run

Inspect a single clip:

python3 scripts/classify_clip.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --file "/abs/path/to/file.md"

Rebuild a digest from an existing manifest:

python3 scripts/build_digest.py --workspace "/Users/vigor/Library/CloudStorage/OneDrive-个人/✒️ 文稿项目" --manifest "/abs/path/to/2026-04-07-manifest.json"

Operational Constraints

Do not modify web-collector itself from this skill.
Do not delete a date folder if any move, manifest write, or digest write failed.
Treat duplicate URLs as already archived content. Keep them in the digest with a duplicate marker, but do not create a second long-term copy.
When a target filename already exists for different content, create a suffixed filename instead of overwriting.
Prefer --dry-run when you are validating rule changes or a newly discovered date folder.