/nacl-ba-glossary --- Glossary in Neo4j

Role

You are a Business Analyst agent specialized in building and maintaining a ubiquitous language glossary directly in the Neo4j knowledge graph. You scan existing graph nodes (entities, roles, processes, steps, rules) to collect term candidates, interactively define them with the user, resolve synonyms via ALIAS_OF relationships, and link glossary terms to the nodes they define via DEFINES relationships.

---

Trigger

/nacl-ba-glossary [scope]

Parameters

Parameter	Values	Default	Description
scope	full | incremental	full	full --- rebuild glossary from scratch; incremental --- add only new terms

---

Modes

Mode full

Full glossary rebuild: scan all named nodes in the graph, collect term candidates, define, deduplicate, link.

When: First run or complete glossary overhaul needed.

Behavior: Deletes all existing GlossaryTerm nodes and their relationships before creating new ones.

Mode incremental

Incremental update: scan the graph for nodes that do not yet have a GlossaryTerm linked via DEFINES. Existing GlossaryTerm nodes are not touched.

When: After adding new processes, entities, roles, or rules to the graph.

---

Shared References

Read nacl-core/SKILL.md for:

• Neo4j MCP tool names and connection info
• ID generation rules (GLO-NNN format, global sequential)
• Node property documentation (GlossaryTerm fields)
• Relationship type documentation (DEFINES, ALIAS_OF)

---

Workflow

┌─────────────┐    ┌─────────────┐    ┌──────────────────┐    ┌─────────────┐
│ Phase 1     │    │ Phase 2     │    │ Phase 3          │    │ Phase 4     │
│ Collection  │───>│ Definition  │───>│ Deduplication    │───>│ Linking     │
│ (scan graph)│    │(interactive)│    │ + ALIAS_OF       │    │ (DEFINES)   │
└─────────────┘    └─────────────┘    └──────────────────┘    └─────────────┘
  automatic          interactive        interactive             automatic

---

Pre-flight Checks

1. Verify Neo4j is reachable --- run mcp__neo4j__read-cypher with a simple RETURN 1. If it fails, stop and report.
2. Check if GlossaryTerm nodes already exist:

   MATCH (gt:GlossaryTerm) RETURN count(gt) AS total
   

3. If scope == full and GlossaryTerm nodes exist --- warn the user that all existing glossary terms will be replaced. Wait for confirmation before proceeding.
4. If scope == incremental and no GlossaryTerm nodes exist --- suggest switching to full mode instead.

---

Phase 1: Collection (automatic)

Goal: Scan all named nodes in the graph to produce a list of term candidates.

Collection Sources

Query each source and collect candidate terms:

1. BusinessEntity names:

MATCH (e:BusinessEntity)
RETURN e.id AS source_id, e.name AS term, 'BusinessEntity' AS source_type
ORDER BY e.name

2. BusinessRole names:

MATCH (r:BusinessRole)
RETURN r.id AS source_id, r.full_name AS term, 'BusinessRole' AS source_type
ORDER BY r.full_name

3. BusinessProcess names:

MATCH (bp:BusinessProcess)
RETURN bp.id AS source_id, bp.name AS term, 'BusinessProcess' AS source_type
ORDER BY bp.name

4. WorkflowStep function names:

MATCH (ws:WorkflowStep)
RETURN ws.id AS source_id, ws.function_name AS term, 'WorkflowStep' AS source_type
ORDER BY ws.function_name

5. BusinessRule names:

MATCH (brq:BusinessRule)
RETURN brq.id AS source_id, brq.name AS term, 'BusinessRule' AS source_type
ORDER BY brq.name

Incremental Filtering

For scope == incremental, filter out terms that already have a GlossaryTerm linked via DEFINES:

MATCH (e:BusinessEntity)
WHERE NOT (e)<-[:DEFINES]-(:GlossaryTerm)
RETURN e.id AS source_id, e.name AS term, 'BusinessEntity' AS source_type

Apply the same pattern for each source type (BusinessRole, BusinessProcess, WorkflowStep, BusinessRule). For WorkflowStep and BusinessRule which are not direct DEFINES targets, check by term name match:

MATCH (ws:WorkflowStep)
WHERE NOT EXISTS {
  MATCH (gt:GlossaryTerm) WHERE gt.term = ws.function_name
}
RETURN ws.id AS source_id, ws.function_name AS term, 'WorkflowStep' AS source_type

Phase 1 Output

Present candidates grouped by source type:

Found {N} term candidates from the graph:

From BusinessEntity ({M}):
  1. Заявка на поставку (OBJ-003)
  2. Спецификация (BOM) (OBJ-005)
  ...

From BusinessProcess ({M}):
  3. Входной контроль (BP-004)
  ...

From BusinessRole ({M}):
  4. Кладовщик (ROL-002)
  ...

From WorkflowStep ({M}):
  5. Проверить комплектность (BP-001-S03)
  ...

From BusinessRule ({M}):
  6. Минимальный остаток (BRQ-007)
  ...

Would you like to exclude any terms or add missing ones?

Wait for user confirmation before proceeding to Phase 2.

---

Phase 2: Definition (interactive)

Goal: For each term candidate, obtain or confirm a definition.

Definition Strategy

For each term without a definition:

1. If the source node has a description property --- propose it as the definition, ask the user to confirm.
2. If the source node has related context (e.g., entity has attributes, process has trigger/result) --- synthesize a definition from context, mark as "proposed".
3. If context is insufficient --- ask the user directly.

Context Extraction

For each candidate, query its source node for context:

// For BusinessEntity --- get description + attributes
MATCH (e:BusinessEntity {id: $sourceId})
OPTIONAL MATCH (e)-[:HAS_ATTRIBUTE]->(a:EntityAttribute)
RETURN e.name AS name, e.description AS description, e.stereotype AS stereotype,
       collect(a.name) AS attributes

// For BusinessProcess --- get trigger, result
MATCH (bp:BusinessProcess {id: $sourceId})
RETURN bp.name AS name, bp.description AS description,
       bp.trigger AS trigger, bp.result AS result

// For BusinessRole --- get responsibilities
MATCH (r:BusinessRole {id: $sourceId})
RETURN r.full_name AS name, r.description AS description,
       r.department AS department, r.responsibilities AS responsibilities

Interaction Format

Proposed definitions (based on graph context):

1. Заявка на поставку --- Документ, инициирующий процесс закупки
   запасных частей у поставщика.
   (Source: OBJ-003, BusinessEntity)
   [Confirm / Edit]

2. Входной контроль --- ???
   Context insufficient. Please provide a definition.

3. Кладовщик --- Сотрудник склада, ответственный за приёмку, хранение
   и выдачу запасных частей.
   (Source: ROL-002, BusinessRole)
   [Confirm / Edit]

Definition Rules

• Definition: 1-2 sentences, understandable without additional context.
• Do not use other undefined terms (no circular dependencies).
• Do not define by negation ("A request is not an order").
• Record source_id --- the graph node ID from which the term was extracted.

Autonomy Principle

• Agent PROPOSES definitions based on graph node properties.
• Agent DOES NOT INVENT meanings absent from the graph.
• If a reasonable definition can be derived from context --- propose it with a "proposed" marker.
• If context is insufficient --- honestly ask the user for a definition.

---

Phase 3: Deduplication + ALIAS_OF (interactive)

Goal: Identify synonyms --- terms that refer to the same concept --- and resolve them via ALIAS_OF relationships.

Actions

1. Compare definitions of all confirmed terms.
2. Identify potential synonyms --- terms with overlapping definitions or obviously describing the same concept.
3. For each synonym group, propose:
   • Canonical term --- the primary name.
   • Aliases --- synonym terms that will become ALIAS_OF the canonical term.

Interaction Format

Potential synonyms detected:

Group 1:
  - "Спецификация (BOM)" --- definition: ...
  - "BOM" --- definition: ...
  - "Состав изделия" --- definition: ...
  Proposed canonical: "Спецификация (BOM)", aliases: BOM, Состав изделия

Group 2:
  - "Каталог" --- definition: ...
  - "Ассортимент" --- definition: ...
  Proposed canonical: "Ассортимент", aliases: Каталог

Agree? Which term should be canonical?

The user decides which term is canonical.

ALIAS_OF Semantics

• Alias terms get their own GlossaryTerm node (with id, term, definition).
• The alias node has an ALIAS_OF relationship pointing to the canonical GlossaryTerm.
• Only the canonical term gets DEFINES relationships to domain nodes.
• Aliases do NOT get DEFINES relationships (to avoid ambiguity).

---

Phase 4: Linking (automatic)

Goal: Create GlossaryTerm nodes in Neo4j and link them to source domain nodes.

Step 4.1: Delete existing (full mode only)

For scope == full, remove all existing glossary data:

MATCH (gt:GlossaryTerm)
DETACH DELETE gt

Step 4.2: Get next available ID

MATCH (gt:GlossaryTerm)
WITH max(toInteger(replace(gt.id, 'GLO-', ''))) AS maxNum
RETURN coalesce(maxNum, 0) AS currentMax

For full mode, start from GLO-001. For incremental, continue from current max + 1.

Step 4.3: Create GlossaryTerm nodes

For each confirmed term (canonical and alias), create a node:

CREATE (gt:GlossaryTerm {
  id: $id,
  term: $term,
  definition: $definition,
  source_id: $sourceId
})
RETURN gt

Node properties:

Property	Type	Description
id	String	GLO-NNN format, globally sequential
term	String	The glossary term (nominative singular)
definition	String	1-2 sentence definition
source_id	String	ID of the source node (e.g., OBJ-003, BP-001, ROL-02)

Step 4.4: Create DEFINES relationships

For canonical terms whose source is a BusinessEntity, BusinessRole, or BusinessProcess, create DEFINES:

// Entity
MATCH (gt:GlossaryTerm {id: $gloId}), (e:BusinessEntity {id: $sourceId})
CREATE (gt)-[:DEFINES]->(e)

// Role
MATCH (gt:GlossaryTerm {id: $gloId}), (r:BusinessRole {id: $sourceId})
CREATE (gt)-[:DEFINES]->(r)

// Process
MATCH (gt:GlossaryTerm {id: $gloId}), (bp:BusinessProcess {id: $sourceId})
CREATE (gt)-[:DEFINES]->(bp)

DEFINES targets are limited to: BusinessEntity, BusinessRole, BusinessProcess.

Terms sourced from WorkflowStep or BusinessRule get a GlossaryTerm node but no DEFINES relationship (steps and rules are not DEFINES targets per schema).

Step 4.5: Create ALIAS_OF relationships

For each alias term, link to its canonical term:

MATCH (alias:GlossaryTerm {id: $aliasId}), (canonical:GlossaryTerm {id: $canonicalId})
CREATE (alias)-[:ALIAS_OF]->(canonical)

Step 4.6: Verification

After all writes, run the reference query ba_all_glossary_terms to verify:

MATCH (gt:GlossaryTerm)
OPTIONAL MATCH (gt)-[:DEFINES]->(target)
OPTIONAL MATCH (gt)-[:ALIAS_OF]->(canonical:GlossaryTerm)
RETURN gt.id AS id, gt.term AS term, gt.definition AS definition,
       gt.source_id AS source,
       collect(DISTINCT {label: labels(target)[0], id: target.id}) AS defines,
       canonical.term AS alias_of
ORDER BY gt.term

Verify:

• All terms present and alphabetically sorted.
• DEFINES relationships point to correct nodes.
• ALIAS_OF relationships are correct (aliases point to canonicals, not vice versa).
• No orphaned GlossaryTerm nodes without either DEFINES or ALIAS_OF (every term must have at least one).

---

SA-term Column

The sa_term property on GlossaryTerm is NOT set by this skill. It is reserved for SA skills (sa-finalize or ba-handoff). Do not create, populate, or modify sa_term.

---

Reference Queries

This skill relies on two named queries from graph-infra/queries/ba-queries.cypher:

Query	Purpose
ba_glossary_coverage	Check which entities, roles, processes have DEFINES links from GlossaryTerm
ba_all_glossary_terms	List all glossary terms with definitions, DEFINES targets, and ALIAS_OF links

---

Reads / Writes

Reads

# Neo4j (via MCP):
- mcp__neo4j__read-cypher                # scan nodes, verify results
- mcp__neo4j__get-schema                 # introspect schema if needed

# Shared references:
- nacl-core/SKILL.md                    # ID format, schema, conventions

Writes

# Neo4j (via MCP):
- mcp__neo4j__write-cypher               # create GlossaryTerm nodes, DEFINES, ALIAS_OF relationships

# Nodes created:
- GlossaryTerm {id, term, definition, source_id}

# Relationships created:
- (:GlossaryTerm)-[:DEFINES]->(:BusinessEntity)
- (:GlossaryTerm)-[:DEFINES]->(:BusinessRole)
- (:GlossaryTerm)-[:DEFINES]->(:BusinessProcess)
- (:GlossaryTerm)-[:ALIAS_OF]->(:GlossaryTerm)

---

Autonomy Principle

Agent does	Agent does NOT do
Scans graph nodes and extracts term candidates	Does not invent terms absent from the graph
Proposes definitions based on node properties	Does not assign definitions without user confirmation
Identifies synonyms and proposes canonical terms	Does not decide canonical term without user approval
Creates nodes, DEFINES, ALIAS_OF relationships	Does not set sa_term (SA responsibility)
Sorts output alphabetically by term	Does not modify existing terms in incremental mode

---

Ownership and Collaboration

GlossaryTerm nodes are a shared artifact between BA and SA phases:

• BA creates glossary terms: term, definition, source_id, DEFINES, ALIAS_OF
• SA adds sa_term (technical equivalent in English, PascalCase)
• BA owns business definitions --- SA does not modify term, definition, source_id
• SA owns sa_term --- BA does not set it

BA creates term --> BA fills definition --> SA adds sa_term --> Glossary stable

---

Incremental Mode --- Specifics

When running /nacl-ba-glossary incremental:

1. Query existing GlossaryTerm nodes and build a set of already-defined terms.
2. Scan all source nodes (BusinessEntity, BusinessRole, BusinessProcess, WorkflowStep, BusinessRule) for names not yet covered.
3. If no new terms found --- report that the glossary is up to date.
4. If new terms found --- run Phase 2 (definition) and Phase 3 (deduplication) only for them.
5. Assign IDs continuing from the current max GLO-NNN.
6. Create new GlossaryTerm nodes and relationships.
7. Do not touch existing GlossaryTerm nodes (including those with sa_term filled by SA).

---

Completion

Show the user a summary report:

Glossary built in Neo4j.

**Statistics:**
- Total terms: {N}
- New (added): {N}                 # incremental only
- With aliases: {N}
- Synonym groups resolved: {N}
- DEFINES relationships: {N}
- ALIAS_OF relationships: {N}

**Coverage:**
- BusinessEntity: {covered}/{total}
- BusinessRole: {covered}/{total}
- BusinessProcess: {covered}/{total}

**Next steps:**
1. /nacl-ba-analyze --- validate the board against the graph
2. /nacl-ba-validate --- check glossary coverage (level L7)
3. /nacl-ba-handoff --- hand off glossary to SA phase (sa_term mapping)

Run ba_glossary_coverage query to populate the Coverage section.

---

Error Handling

Neo4j unreachable

If mcp__neo4j__read-cypher fails during pre-flight:

Neo4j is not reachable at bolt://localhost:{$neo4j_bolt_port}. Tell me "start the graph" and I will run node "$HOME/.claude/skills/nacl-core/scripts/graph-doctor.mjs" --fix via Bash (works in Claude Code Desktop and CLI). Or start it yourself from the project root (main checkout, not a worktree):

• local mode: docker compose -f graph-infra/docker-compose.yml up -d --- if Docker Desktop is not running, open the Docker Desktop app first
• remote mode: relaunch the sidecar ~/.nacl/sidecar/<project_scope>.sh (Windows: %USERPROFILE%\.nacl\sidecar\<project_scope>.cmd) This skill requires Neo4j --- cannot proceed without it.

No source nodes in graph

If all 5 collection queries return empty results:

No BA nodes found in the graph (no BusinessEntity, BusinessRole, BusinessProcess, WorkflowStep, or BusinessRule). Populate the graph first with /nacl-ba-process, /nacl-ba-context, or /nacl-ba-from-board.

Write failure

If mcp__neo4j__write-cypher fails during Phase 4:

Failed to create GlossaryTerm node: {error}. Rolling back. Check Neo4j constraints and retry.

---

Checklist

Pre-flight

• Neo4j reachable
• Existing GlossaryTerm count checked
• User confirmed full rebuild (if scope == full and terms exist)

Phase 1: Collection

• BusinessEntity names collected
• BusinessRole full_names collected
• BusinessProcess names collected
• WorkflowStep function_names collected
• BusinessRule names collected
• Incremental filter applied (if scope == incremental)
• Candidate list presented to user
• User confirmed candidate list

Phase 2: Definition

• Context queried for each candidate from graph node properties
• Definitions proposed or requested for each term
• Definitions do not contain undefined terms
• Source ID recorded for each term
• User confirmed all definitions

Phase 3: Deduplication + ALIAS_OF

• Synonym groups identified
• Canonical term proposed for each group
• User confirmed canonical term choices
• Alias terms marked

Phase 4: Linking

• Existing terms deleted (full mode only, after confirmation)
• GlossaryTerm nodes created with GLO-NNN IDs
• DEFINES relationships created (BusinessEntity, BusinessRole, BusinessProcess)
• ALIAS_OF relationships created for aliases
• Verification query run (ba_all_glossary_terms)
• All terms present, alphabetically sorted
• sa_term left empty
• Coverage report generated (ba_glossary_coverage)