metadata-naming（元数据命名）

Metadata Naming

Turn loose filename habits into a stable naming standard that is easy to read, sort, parse, and reuse.

Use This Standard

Follow this split by file purpose:

• Use stable identity naming for long-lived entries that are updated in place.
• Use timestamp naming for snapshots, archives, exports, and reports.
• Prefer ASCII, no spaces, and fixed separator rules when filenames will be processed by scripts or moved across systems.

Core Metadata Blocks

Use these blocks in a fixed order when a filename needs richer metadata:

• time
• prefix
• title
• version
• tags
• source_or_author
• note

Not every block is required. Keep only the fields that improve retrieval or automation.

Two Modes

Relaxed mode

Use for human-managed documents where readability matters more than strict parsing.

Conventions:

• Chinese or mixed-language titles are allowed.
• Spaces are allowed when the surrounding system tolerates them.
• Typical visual markers:

- time: (2026-03-11) or (20260311-093500) - version: (v1.2.0) - tags: #tag - source or author: @name - note: &note

Example:

(2026-03-11)Favorites Curator(v0.1.0)#skill#favorites@workspace&initial publish

Strict mode

Use by default for standards, skills, inventories, generated files, syncable folders, and anything a script will read.

Conventions:

• ASCII only
• no spaces
• top-level separator: __
• intra-block separator: - or .
• lowercase slug-style titles unless there is a strong reason not to

General template:

YYYYMMDD[-HHMMSS]__prefix__title__version__tags__source__note.ext

Examples:

20260311__skill__favorites-curator__v0.1.0__favorites.catalog__workspace.md
20260311__repo__openclaw-backup-tool__v0.1.0__backup.tool__github.md
20260311__app__codex__v0.112.0__cli.ai__brew.md

Default Rule For Long-Lived Entries

For catalogs, inventories, and canonical records, prefer stable filenames over timestamped filenames.

Use this template:

<data_type>__<source_name>__<slug>.md

Examples:

skill__workspace__favorites-curator.md
repo__github__openclaw-backup-tool.md
app__brew__codex.md

Use this rule when the content is refreshed in place and the filename should not drift over time.

Default Rule For Snapshots And Reports

Use timestamp-first filenames for time-series artifacts.

Templates:

YYYYMMDD__report__topic.md
YYYYMMDD-HHMMSS__snapshot__topic.json

Examples:

20260311__report__favorites-digest.md
20260311-095914__snapshot__favorites.json

Block Rules

Time

• Use YYYYMMDD for day-level tracking.
• Use YYYYMMDD-HHMMSS for run-level uniqueness.
• Put time first when sort order matters.

Prefix

• Use short taxonomy values such as repo, skill, app, doc, snapshot, report.
• Keep the vocabulary stable once chosen.

Title

• Make this the main identity.
• Prefer short, stable, searchable slugs.
• Move extra description into tags or notes.

Version

• Use semver when available: v0.1.0, v2.3.4.
• Omit the block when versioning is irrelevant.

Tags

• Use compact, low-noise tags.
• Join multiple tags with . or - inside the same block.

Source Or Author

• Use the source system, publisher, owner, or author when it improves retrieval.
• Examples: github, brew, workspace, openclaw, vendor-name.

Note

• Keep it short.
• Do not put long prose into filenames.
• Use only when the note materially changes retrieval value.

Standard Principle

Use fixed-order metadata blocks. Use stable identity filenames for long-lived entries and timestamped filenames for snapshots. Default to ASCII, no spaces, __ between blocks, and - or . inside blocks so names stay sortable, parseable, and portable.

References

Read references/standard.md when you need the normalized standard, examples, and decision rules in a reference-friendly format.