> For the complete documentation index, see [llms.txt](https://docs.decube.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.decube.io/mcp/skills.md). # Skills {% hint style="info" icon="telescope" %} Skills are in **limited preview**. We're iterating on it quickly based on how teams actually use it, so tools, capabilities and availability may change at short notice. For access, questions, or feedback on how it's working for you, reach out at . {% endhint %} A Decube skill is a documented, repeatable workflow. Instead of working out which tools to call and in what order, you run the same expert sequence every time you need it. Each skill is available on two surfaces: * **Your own AI client:** download the skill definition and drop it into Claude, Cursor, or another MCP-compatible client. * **Trusty AI:** add the same skill definition to Decube's in-product assistant. ## Find your skills by role Every skill below is available to everyone. We've grouped them by role only to make this page easier to read, not a rule about who can run what. | Group | What these skills do | | --------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------- | | [**Data analysts**](#for-data-analysts) | Find the right table, confirm what a metric means, and check nothing is broken before you present | | [**Data engineers**](#for-data-engineers) | Know what breaks before you change it (blast radius analysis), catch schema drift, and work incidents and monitor coverage | | [**Data stewards and governance leads**](#for-data-stewards-and-governance-leads) | Track certification, ownership and monitor SLAs across the estate, and get a daily briefing | ## For data analysts For when you're building a report and need to know the numbers are right before anyone else sees them. ### Query-Safe Table Finder * Searches the catalog for tables matching a business topic * Checks certification, significance tier, ownership and quality score on each candidate * Ranks them and names the one to use, plus the ones that look right but aren't **When to use it** * When you don't know which table to query for a topic and don't want to guess from names * When several tables look like they could be the source of truth * Before building anything new on a table you haven't used before {% file src="/files/UvK3qaVZpHB0et4Wh0mi" %} ### Metric Catalog Lookup * Finds where a business metric is actually defined in the catalog * Pulls the column, description, owner and any custom attributes carrying the formula * Flags it when two tables claim the same metric with different definitions **When to use it** * When you need to confirm how a metric is computed before using it * When two dashboards disagree on the same number * When you want the owner to confirm an edge case with {% file src="/files/39evHJN7ixFaKqqRFwvu" %} ### Dashboard Trust Check * Resolves every table feeding a dashboard or report through lineage * Checks each one for open incidents, failing monitors, staleness and certification status * Returns a go/no-go, naming which specific number is at risk **When to use it** * Before presenting a dashboard to stakeholders or the board * When you want to know whether today's numbers can be trusted, not last week's * When something looks off and you need to know if the data underneath moved {% file src="/files/kxVsOKc9cMIcJ0ySJbPN" %} ## For data engineers For when you're changing something in the pipeline and need to know what breaks, or a check has fired and you need to work out what to do about it. ### Pipeline Impact Check * Traces upstream and downstream lineage for a table, column or pipeline * Enriches the downstream list with significance tier, asset type and monitoring coverage * Returns a blast-radius verdict, calling out unmonitored downstream assets as silent-failure risks **When to use it** * Before dropping, renaming or retyping a column on a shared table * Before any migration or refactor that touches something other teams read * When you need to know who to warn, not just what depends on what {% file src="/files/IsjcwTXs8w1zzaGcZtNP" %} ### Schema Drift Watch * Diffs the two most recent profiling runs for a table, schema or source * Reports new columns, dropped columns and type changes * Flags drift that lands on a Tier1 asset or breaks an existing monitor **When to use it** * When you suspect an upstream source changed shape without telling anyone * As a recurring check on the sources you can't afford to have drift silently * When a monitor started failing and you want to know if the table changed underneath it {% file src="/files/m60yOTp5ZCd6rmvi35W0" %} ### Monitor Recommendation * Reads an asset's profiler statistics and catalog metadata * Proposes a ranked list of the monitors best suited to it * Creates the ones you approve **When to use it** * After connecting a new source or schema with little monitor coverage * Whenever you want a starting set of monitors on an asset instead of building from scratch * To improve the data quality coverage of your assets {% file src="/files/UUhkTI7Jq0FNXOS1LJNo" %} ### Incident Remediation * Reads the incident, the monitor that fired, and its run history to judge whether the data broke or the rule is miscalibrated * Walks lineage upstream for the root cause and downstream for the blast radius * Returns a remediation plan: the diagnosis, the likely root cause, the consumers affected, ranked next steps with named owners, and how the fix gets verified **When to use it** * When you need a remediation plan for an incident * Alongside [Fix with Trusty AI](/trusty-ai/fix-incident-with-trusty.md) on the Incident Details page * When a monitor keeps firing and you need to know whether to fix the data or retune the check {% file src="/files/qjKXisVCIh8ufjHcNNdW" %} ## For data stewards and governance leads For when the question is coverage across the estate rather than any one table. These report on the gaps. ### Tier 1 Certification Sweep * Resolves the Significance Tier and Certification attributes fresh, then sweeps every Tier1 asset in scope * Reports how many are Verified, Pending, and have no certification set at all * Lists every non-Verified Tier 1 asset grouped by source, with its owner **When to use it** * Ahead of an audit or a board review, when coverage numbers need to be real * When you want to know which critical assets nobody has started certifying * On a recurring cadence, with the scope baked in {% file src="/files/5Z36qUvufWPOP5fNNVcS" %} ### Quality SLA Tracker * Pulls every monitor in scope with its run history and current status * Classifies each as disabled, stale, failing or healthy against your SLA bar * Leads with disabled and failing monitors on Tier1 assets, the gaps nobody is watching **When to use it** * When you need to know whether your monitors are actually running, not just configured * To find monitors that were disabled once and never turned back on * As a recurring compliance check against a stated quality SLA {% file src="/files/sB3mHRH4DuxqKyPSwreF" %} ### Steward Morning Brief * Pulls open incidents, the certification queue, Tier1 ownership gaps and quality degradation in one pass * Renders a single self-contained HTML page, with every asset and incident linking back into Decube * Includes a board-ready view of estate health broken down by source **When to use it** * As a daily first-thing view of what needs a steward's attention * When you need a CIO or board update on estate health that isn't a screenshot of a dashboard * On a schedule, so the briefing is waiting for you each morning {% file src="/files/OiDmop1Qtd5AvoW1c3YK" %} ## Using skills in your own AI client Download the skill definition and add it to your MCP-compatible client, then run it against Decube's tools. You'll need the Decube MCP server connected first: see [How to Setup Decube MCP](/mcp/how-to-setup-decube-mcp.md) if you haven't set that up yet. ## Using skills in Trusty AI The same skill definitions work inside [Trusty AI](/trusty-ai/trusty-ai.md). Add the skill to your Trusty interface and it runs against the tools Trusty already has, no separate client setup required.