Altertable Ask
Central entry point for Altertable skills. Every Altertable task starts here. Match the user query to the best available skill and hand off.
Procedure
- Read the user's query.
- If using Altertable MCP tools, call
initialize before any other Altertable tool so organization, environment, and knowledge context are correct.
- Match the query against the routing table below, then apply the rules.
- Activate the matched skill in your harness, passing the original query through so the matched skill has full context.
- If the query is Altertable-related but no skill matches with confidence, use
understand-platform to orient the user.
For data questions, route to explore-data when schema is unclear; otherwise route to query-lakehouse.
Routing Table
| Skill |
When to route |
explore-data |
Discover what data exists: catalogs, schemas, tables, columns, semantic models |
query-lakehouse |
Answer questions that require querying lakehouse data using SQL |
analyze-funnels |
Build or analyze a step-by-step conversion flow (drop-off between ordered events) |
analyze-web-traffic |
Web analytics: pageviews, sessions, traffic sources, UTM, device, country breakdowns |
analyze-insights |
Interpret an existing Insight or visualization the user is looking at |
build-segments |
Define or compare user cohorts by properties (not step-based) |
forecast-timeseries |
Project future metric values or detect whether a change is within normal range |
instrument-product-analytics |
Add or change event tracking, identification, traits, consent, session reset, or identity aliasing in an application |
query-product-events |
Query or analyze existing product events, properties, identities, traits, or instrumentation delivery |
create-insights |
Create a new Insight that will be saved and visible to users |
create-discoveries |
Create a discovery from a meaningful finding that should notify users and enter review |
decide-actions |
Decide which insight or task type to use, or whether to create / update / skip a discovery |
manage-discoveries |
Review, approve, or reject existing discoveries and process user feedback on them |
configure-tasks |
Set up a scheduled AI task (anomaly detection, forecast, monitor) that runs on a cron |
use-memory |
Search or create agent memories; create organization knowledge only when the user explicitly asks |
evaluate-skills |
Review or author agent skills themselves (skill structure, spec, quality) |
understand-platform |
Explain Altertable concepts, architecture, or how agents work |
When a skill is added, renamed, or removed from this repository, update this table in the same change.
Routing Rules
- Single best match: pick one skill. Do not fan out.
- Prefer the narrower skill: when two skills could match, prefer the more specific one.
- Data-first when intent is vague: if the user wants to analyze data but does not specify how, start with
explore-data.
- Fallback for unknown queries: if nothing matches with confidence, route to
understand-platform.
- Pass context through: hand the original query to the matched skill.
- Never invent a skill: only invoke skills that are actually installed.
- Clarify before routing: if the query could reasonably mean different things, propose the most likely directions and let the user choose.
- Separate writes from analysis: route application or ingestion changes to
instrument-product-analytics; route questions about stored event data to query-product-events.
1---2name: ask3description: Routes user queries to the best-fit Altertable skill. Use when unsure which Altertable skill applies to a request.4---56# Altertable Ask78Central entry point for Altertable skills. Every Altertable task starts here. Match the user query to the best available skill and hand off.910## Procedure11121. Read the user's query.132. If using Altertable MCP tools, call `initialize` before any other Altertable tool so organization, environment, and knowledge context are correct.143. Match the query against the routing table below, then apply the rules.154. Activate the matched skill in your harness, passing the original query through so the matched skill has full context.165. If the query is Altertable-related but no skill matches with confidence, use `understand-platform` to orient the user.1718For data questions, route to `explore-data` when schema is unclear; otherwise route to `query-lakehouse`.1920## Routing Table2122| Skill | When to route |23|-------|---------------|24| `explore-data` | Discover what data exists: catalogs, schemas, tables, columns, semantic models |25| `query-lakehouse` | Answer questions that require querying lakehouse data using SQL |26| `analyze-funnels` | Build or analyze a step-by-step conversion flow (drop-off between ordered events) |27| `analyze-web-traffic` | Web analytics: pageviews, sessions, traffic sources, UTM, device, country breakdowns |28| `analyze-insights` | Interpret an existing Insight or visualization the user is looking at |29| `build-segments` | Define or compare user cohorts by properties (not step-based) |30| `forecast-timeseries` | Project future metric values or detect whether a change is within normal range |31| `instrument-product-analytics` | Add or change event tracking, identification, traits, consent, session reset, or identity aliasing in an application |32| `query-product-events` | Query or analyze existing product events, properties, identities, traits, or instrumentation delivery |33| `create-insights` | Create a new Insight that will be saved and visible to users |34| `create-discoveries` | Create a discovery from a meaningful finding that should notify users and enter review |35| `decide-actions` | Decide which insight or task type to use, or whether to create / update / skip a discovery |36| `manage-discoveries` | Review, approve, or reject existing discoveries and process user feedback on them |37| `configure-tasks` | Set up a scheduled AI task (anomaly detection, forecast, monitor) that runs on a cron |38| `use-memory` | Search or create agent memories; create organization knowledge only when the user explicitly asks |39| `evaluate-skills` | Review or author agent skills themselves (skill structure, spec, quality) |40| `understand-platform` | Explain Altertable concepts, architecture, or how agents work |4142When a skill is added, renamed, or removed from this repository, update this table in the same change.4344## Routing Rules45461. **Single best match**: pick one skill. Do not fan out.472. **Prefer the narrower skill**: when two skills could match, prefer the more specific one.483. **Data-first when intent is vague**: if the user wants to analyze data but does not specify how, start with `explore-data`.494. **Fallback for unknown queries**: if nothing matches with confidence, route to `understand-platform`.505. **Pass context through**: hand the original query to the matched skill.516. **Never invent a skill**: only invoke skills that are actually installed.527. **Clarify before routing**: if the query could reasonably mean different things, propose the most likely directions and let the user choose.538. **Separate writes from analysis**: route application or ingestion changes to `instrument-product-analytics`; route questions about stored event data to `query-product-events`.
altertable-ai