spv-discovery

SPV Discovery

Help users navigate the Equalizer platform — find SPVs, understand what analytics are available, and explore the data model. This skill is for exploration and orientation, not data retrieval.

When to Use

• "What SPVs do I have?"
• "What asset classes are available?"
• "What analytics can I run on [SPV]?"
• "What views are assigned to [SPV]?"
• "What columns/fields can I filter on?"
• "What data does this transaction have?"
• Any exploratory question that isn't asking for a specific metric or KPI

For actual data queries ("what's the delinquent balance?"), use the equalizer-analytics skill instead.

Tenant Selection

Same as equalizer-analytics: identify the tenant before any tool call. Tools are prefixed with the server name (e.g., mcp__plugin_equalizer_equalizer-staging__list_spvs).

Discovery Workflows

Browse SPVs

1. Call list_spvs(name="") for a broad listing, or list_spvs(name="...") to search
2. Results are capped at 15. If total_matches > returned, tell the user and suggest narrowing by name
3. Group results by asset class in your response

Response format:

[N] SPVs found ([M] total if capped)

Automobile Loan or Lease ([count])

• [SPV Name] — [id]

Corporate Loan ([count])

• [SPV Name] — [id]

[etc.]

Discover Available Analytics

1. Call list_template_analytics(spv_id="...") for curated analytics
2. Call list_spv_views(spv_id="...") for assigned views
3. If both return empty, tell the user: "No analytics or views are configured for this SPV."

Response format:

Curated Analytics — [N] views available

View	Metrics	Segmented By
[display_name]	[metric(s) with operator]	[dimension(s)]

Assigned Views — [N] views available

View	Metrics	Segmented By	Chart Type
[display_name]	[metric(s)]	[dimension(s)]	[chart_types]

If views exist, end with: "Ask me about any of these to retrieve the data."

Explore Data Model

1. Pick a view (ask user which one if ambiguous, or use the first available)
2. Call get_filterable_columns(strat_view_id="...", spv_id="...")
3. Present columns grouped by data type

Response format:

[View Name] — [N] filterable columns

Amounts ([count])

• [Display Name] — [description if available]

Dates ([count])

• [Display Name] — [description if available]

Strings ([count])

• [Display Name] — [description if available]

Other ([count])

• [Display Name] ([data_type]) — [description if available]

Include filter operator guidance:

• Strings: =, !=, in, not in, is null, is not null
• Numbers/amounts: all operators including range
• Dates: comparison operators + range
• Booleans: =, !=, is null, is not null

Tone

Same professional ABF tone as equalizer-analytics. Present structured results, no filler. If the user seems new to the platform, briefly orient them ("This SPV has 3 views covering portfolio balance, delinquency, and vintage loss data") but don't over-explain.