Edge types

Edges are typed, directional, and weighted. Every edge is written from the source node’s perspective: - {type: resolves, to: issue-webhook-retries} on a decision means “this decision resolves that issue”. Weights control how far the briefing compiler follows a relationship — high-weight edges decay slowly across hops, which is what makes structural traversal beat plain similarity search.

Prefer one precise high-weight edge over three relates-to.

Lineage and dependency edges

Edge	Weight	Meaning (from the source node)
supersedes	1.0	This node replaces the target; the target is stale.
constrained-by	1.0	The target limits what this node may do.
governed-by	0.95	The target is the norm or policy this node falls under.
derived-from	0.9	This node was produced from the target.
decided-in	0.9	The choice in this node was made in the target.
resolves	0.9	This node fixes or closes the target — the edge the resolver gate requires.
blocks	0.7	The target cannot proceed until this node does.
answers	0.7	This node answers the target question, pulling the answer through the asker’s next briefing.
relates-to	0.5	Weak association.
mentions	0.5	Weakest association.

People and routing edges

Edge	Weight	Meaning (from the source node)
assigned	0.5	This work is assigned to the target person or agent — the explicit-routing edge. An agent target may carry a profile: override.
reviewed-by	0.5	The target person reviewed and approved this node; counts toward a policy quorum.
changes-requested-by	0.5	The target person reviewed this node and requested changes — not an approval.
stewards	0.4	The source person stewards the target area, spec, or norm — the question-routing key.
review-requested	0.3	A review of this node is requested of the target person; surfaces in their queue.
routed-to	0.3	This question was routed to the target person for answering.

The three review edges are mutually exclusive per node-and-person pair: the edge type is that reviewer’s current verdict, and writing one flips any sibling review edge to the same person in place. A reviewer turns a pending review-requested into reviewed-by with a single edge write.

Structural identity edges

Low-weight by design: they bind identities together without pulling owners and groupings into every work neighborhood.

Edge	Weight	Meaning (from the source node)
member-of-org	0.3	The source person is a member of the target organization.
grouped-under	0.3	The source repo’s home project grouping.
owned-by	0.3	The source agent is owned by the target person.
uses-profile	0.3	The source agent’s default dispatch profile.

Workflow and provenance edges

Edge	Weight	Meaning (from the source node)
performs	0.8	This run is an execution of the target workflow.
triggered-by	0.7	This run was triggered by the target node or event.
compiled-for	—	Briefing to its task or query; provenance only, no traversal weight.
shaped-by	—	Briefing to the corrections applied to it; provenance only.

Inverse labels and aliases

Write edges in the canonical direction above, but the write path accepts two normalized forms:

• Inverse labels — the same edge read from the target’s side. Writing {type: blocked-by, to: issue-webhook-retries} on task-carrier-rollout is flipped and stored on the issue as {type: blocks, to: task-carrier-rollout}. Registered inverses: blocked-by → blocks, answered-by → answers, superseded-by → supersedes, groups → grouped-under, owns → owned-by, has-org-member → member-of-org.
• Aliases — same-direction synonyms renamed in place: related-to → relates-to, derives-from → derived-from, supercedes → supersedes, approved-by → reviewed-by.

Unknown edge types beyond this vocabulary are rejected, not silently defaulted. Hand-written nodes should use the canonical forms.

Weights are registry data

The table above shows the seed pack’s weights. Like everything else in the ontology, edge types and their weights live in schema nodes, so a team can add an edge type or retune a weight by writing a schema node rather than waiting for a release. spor schema --edges prints the live table for your graph. View-layer schemas may add their own structural edges — for example, a workspace mirrors its lens slots with low-weight composes edges.