document-classify

name	document-classify
description	Apply a taxonomy label, sensitivity, and retention bucket to a document.
inputs	[{"name":"document_id","type":"string","required":true,"description":"UUID of the Document to classify."},{"name":"taxonomy","type":"string","required":true,"description":"Target taxonomy namespace (e.g. \"iso-15489\", \"internal\")."},{"name":"force","type":"boolean","required":false,"description":"Overwrite an existing classification if true. Defaults to false."}]
outputs	[{"name":"label","type":"string","description":"Assigned taxonomy label."},{"name":"sensitivity","type":"string","description":"Sensitivity tier assigned (e.g. PUBLIC, INTERNAL, CONFIDENTIAL)."},{"name":"retention_bucket","type":"string","description":"Retention period bucket assigned (e.g. 3Y, 7Y, PERMANENT)."}]
tools	["db-query","llm-classify"]

document-classify

When to use

Invoke this skill when a document has been ingested but lacks a classification, or when the user asks to re-classify under a new taxonomy. Run before document-redact so sensitivity tier is known, and before document-workflow so approval routes can be chosen by label.

How it works

• Fetch Document.summary and up to 3 Chapter.summary rows via db-query.
• Send the summaries plus the taxonomy namespace to llm-classify and receive a structured label, sensitivity, and retention bucket.
• Patch the Document row with the results via db-query.
• If DocumentInstance rows exist (derived copies), propagate the sensitivity tier.

Audit

Emits one ExecutionHistory row with event_type=DOCUMENT_EDITED and actor_type=agent.