Metadata Completeness Auditor

Metadata Completeness Auditor

Purpose

Missing or inconsistent metadata is the single most common reason SharePoint search, filtering, views, and retention stop working. This skill measures the metadata quality of a library (or list), flags every item with gaps, and produces a color-coded HTML dashboard the team can act on. It is strictly read-only: it inspects the schema and reads items, but never creates columns, writes to items, or changes the list in any way.

Trigger Phrases

Activate this skill when the user says any of the following (or close variations):

• "metadata audit" / "audit metadata"
• "metadata completeness" / "metadata health" / "metadata quality"
• "missing metadata" / "which items are missing metadata"
• "check metadata" / "metadata gap report"
• "find blank columns" / "find empty fields"

Inputs & Scope

Determine the reporting scope from the user's request and the current context, in this order:

1. Selected items — if the user has items selected or says "these/this", audit the library that contains the selection.
2. Named library or list — if the user names one (e.g., "audit the Contracts library"), resolve it on the current site.
3. Current library — if the user is in a library and gives no other scope, audit it.
4. If no scope can be resolved, ask the user which library or list to audit. Do not guess.

Only audit content the current user can already see. Never invent items or values.

Steps

Step 1 — Resolve the target list and read its schema

Use get_list_schema on the resolved library/list to enumerate its columns. Record for each column: internal name, display name, type (text, choice, number, date, person, managed metadata, multi-value, etc.), and whether it is marked required.

Step 2 — Decide the "required set" of columns to audit

Choose which columns count toward completeness, in this order of preference:

1. Explicit user request — if the user names specific columns (e.g., "check Owner and Review Date"), audit exactly those.
2. Columns marked required in the schema — if any exist, default to auditing those.
3. All user-created content columns — if nothing is marked required, audit every non-system content column.

Always exclude system/internal and label columns from the audit and from gap counts, e.g.: ID, Title, LinkTitle, FileLeafRef (Name), ContentType, Modified, Created, Author, Editor, _ModerationStatus, FileSizeDisplay, thumbnail/compliance/checkout columns, and other underscore-prefixed system fields. (You still read Editor/Author for owner grouping and FileLeafRef/Title as the item label — just don't score them.)

Do not count Title/Name toward completeness: it is the item's label, is almost always populated, and scoring it inflates the completeness % and hides truly-empty items. Only audit it if the user explicitly asks to include it.

Briefly tell the user which columns you are auditing before proceeding. If the required set is ambiguous and the user gave no guidance, state the set you selected and continue; only pause to ask if the schema is unreadable.

Step 3 — Read every item

Use list_items with field projection limited to the audited columns plus Editor/Author (for owner grouping) and FileLeafRef/Title (for item labels). Page through the whole library — do not stop at the first page. For very large libraries, process in batches and keep running totals. If a read fails or returns partial data, record that plainly in the report's limitations section rather than filling gaps with assumptions.

Step 4 — Evaluate each item

For each item, evaluate each audited column and classify the value as Missing when it is:

• empty or null,
• whitespace-only,
• an empty multi-value / empty person / empty managed-metadata field, or
• a placeholder value (case-insensitive), including: N/A, NA, TBD, TBC, -, --, ., none, null, n\a, pending, choose an option, select..., enter value, xxx, test, and the column's own default choice when it clearly means "not set".

If the user provides their own placeholder list, use it instead of (or in addition to) the defaults, and note that in the report.

Then compute:

• Per-item completeness % = present columns / audited columns.
• Item verdict: Complete (all audited columns present), Partial (some present), or Empty (none present).
• Per-column fill rate = items with a present value / total items, tracked across the run.
• Overall completeness % = total present values across every audited cell ÷ (number of items × number of audited columns). This is a cell-level fill rate — not the share of Complete items and not the average of the per-item percentages. Report it as a whole number.
• Per-owner rollup = completeness grouped by Editor (fallback Author).