---
name: knowledge-rule
description: Choose a canonical home and capture messages, paragraphs, or notes in a knowledge storage using its local rules. Use for filing knowledge or deciding where it belongs, not ordinary Markdown formatting or repository documentation edits.
---

# Knowledge rules and capture

If no `meta:*` skill is available in this session, warn once; continue with available
local instructions.
Read the relevant storage map and destination rules. The map describes purpose,
inbox (what and how), and relationships; lifecycle and maintenance live locally.
If no destination inbox is defined, retain the source and resolve that gap
instead of inventing a folder. Do not inspect unrelated storages to file one item.

## Ownership

- One canonical home per object. Elsewhere, keep links and local context.
- Separate information by purpose and audience. A research question and a
  personal task to investigate it have different lifecycles. Professional and
  personal information about one person can belong in different stores.
- A paragraph can remain a block. Extract it when it needs its own backlinks,
  status, or lifecycle, not because the containing note is long.
- Use human-readable names and existing identifiers; retain useful aliases
  when renaming. No universal ID or metadata schema is required.
- State belongs in fields; tags are cross-cutting classifications. Report
  conflicts with existing local conventions as proposed migrations.
- Sources are durable; views are computable. An authored overview can contain
  knowledge that a generated index cannot replace.

## Capture

Save the coherent unit with provenance: its source and exact message, block,
or page when available. Use local conventions for dates and metadata; do not
invent a source identifier. Reuse an existing capture of the same source unit
in that destination; a partial capture may still need its missing excerpt.

Capture into the prescribed inbox without doing the later processing session,
unless the user requests processing or names an existing destination paragraph.
Distinguish dated source artifacts, which survive processing, from temporary
raw inbox material. Preserve meaning and provenance when extracting a paragraph.

For crossing storage boundaries, read [cross-boundary](../cross-boundary/SKILL.md).
For later processing, use [process-inbox](../process-inbox/SKILL.md).