Skip to main content
sourceSource: corpus/docs/adrs/0061-intake-artifact.mdModified: 2026-06-23

type: adr id: adr-0061 status: accepted created: 2026-06-11 updated: 2026-06-11

ADR-0061 — The intake artifact and the Pull step

Context

Work usually originates in an external tracker (Jira, Linear, GitHub, Notion). Specs interpret; without a captured original, nobody can later check what was actually asked, and upstream edits silently orphan the spec. Structured intake is mainstream practice in the trackers themselves (issue forms/templates).

Decision

  1. Intake (type: intake) is the thinnest possible snapshot: frontmatter source, url, captured (date) plus the upstream content pasted verbatim — never edited, never interpreted. The spec cites it in sources[].
  2. Intake is recommended when work originates in an external tool, never required. It is not a spec parent with its own pipeline: it is preserved input. PRDs, research, and bug reports remain their own (advanced) artifacts.
  3. Pull is the loop's first step: capture the upstream item into intake/, by copy-paste today. Tracker connectors are a future-CLI concern (corpus pull); this repo ships none.

Alternatives considered

AlternativeWhy weaker
Spec links the tracker URL onlyUpstream items get edited/deleted; the interpretation loses its anchor
Rich intake schema (fields, labels, comments)Re-invents the tracker; verbatim paste is the honest, zero-ceremony floor

Consequences

Positive: ticket→spec provenance for free. Negative: one more (tiny) file per externally-sourced item.

Status

Accepted. Refines ADR-0030; partially supersedes ADR-0053's deferral of an intake directory (the no-connectors clause is reaffirmed).

Propagation

Template (intake), docs/02/10, workspace tree (0060), conformance fixture, future-cli page.

Ready to run the loop on your own repo? Get started — copy the kit and write your first spec.