autolink-entities

name	autolink-entities
description	Scan an entity's summary/metadata and cross-reference the graph to suggest new edge relationships (FACED, INFLUENCED_BY, LED, RIVALED, MENTORED, etc.) in Apache AGE. Use when the Knowledge Graph agent must enrich the Human Story Graph after a new entity or source is added.
metadata	{"agent":"knowledge-graph","source":"Project Atlas Agent Skills Manifest §3","layer":"memory"}

autolink-entities

Propose and write new graph edges for an entity.

When to use

• A new personality/source was added and its relationships are not yet modeled.
• Periodic graph-enrichment pass.

Relationship types (spec §4.5)

INSPIRED, RIVALED, FAILED, REINVENTED, LED, MENTORED, INFLUENCED, ACQUIRED e.g. (RatanTata)-[:LED]->(TataGroup).

Function signature (manifest contract)

def autolink_entities(entity_id: str) -> list:
    """Cross-reference the graph and return suggested new edges for entity_id."""

Inputs / Outputs

• Input: entity_id.
• Output: list of suggested edges {from, type, to, confidence}.

Implementation

Call the temporal-bridge MCP tool start_workflow("autolinkEntitiesWorkflow", "knowledge-graph", [{"entity_id": entity_id}]), then poll get_workflow_result(workflow_id). Reads the entity + its existing relationships rows and nearby candidates (via query-semantic-nodes) from the live Railway Postgres graph, then one LLM call proposes typed edges restricted to the vocabulary above (server/src/modules/graph/graph.service.ts's autolinkEntities). Returns suggestions only -- does not write to relationships itself; a human/Curator must approve and persist them separately, per the existing rule below.

Backend dependency

• Implemented as plain Postgres entities/relationships tables (live on Railway), not the Apache AGE extension this skill's description originally assumed -- the spec (sys_graph.spec.md) only requires node/edge CRUD + vector lookup + edge suggestion, which this schema already satisfies.

Model

deepseek-direct/deepseek-chat to read a summary and propose typed edges; the human/Curator approves before persisting. Output schema-only JSON.