Update to Data Bloo

{
  "name": "data-bloo-channel-analysis",
  "description": "Analyze one marketing, search, social, local, or ecommerce data source connected to Data Bloo. Use when a user selects a source such as Facebook Ads and asks for a performance analysis, trends, strongest and weakest results, or recommendations based on that source's data. Do not use for cross-channel comparisons or for creating, editing, pausing, publishing, or deleting external data.",
  "included_files": [],
  "skill_md_contents": "---\nname: data-bloo-channel-analysis\ndescription: Analyze one marketing, search, social, local, or ecommerce data source connected to Data Bloo. Use when a user selects a source such as Facebook Ads and asks for a performance analysis, trends, strongest and weakest results, or recommendations based on that source's data. Do not use for cross-channel comparisons or for creating, editing, pausing, publishing, or deleting external data.\n---\n\n# Data Bloo Channel Analysis\n\nUse the Data Bloo MCP to analyze one connected data source and one selected account. All Data Bloo tools are read-only.\n\n## Trigger conditions\n\nUse this skill when the user asks to:\n\n- Analyze one selected marketing, search, social, local, or ecommerce source.\n- Review the performance of a source such as Facebook Ads.\n- Identify trends, strongest and weakest campaigns, products, pages, posts, keywords, or other entities within one source.\n- Summarize performance and provide practical recommendations based on connected data.\n- Produce a detailed report for one source.\n\nDo not use this skill when the user asks to:\n\n- Compare two or more different channels or connectors.\n- Create, edit, pause, publish, or delete campaigns, accounts, ads, dashboards, reports, or source data.\n- Buy, upgrade, or manage a Data Bloo subscription.\n- Give general marketing advice that does not require the user's connected Data Bloo data.\n- Analyze a source that is not returned by `list_data_sources`.\n\n## Workflow\n\n### 1. Confirm the source\n\nIdentify the single data source the user wants to analyze.\n\nIf the source is not specified:\n\n1. Call `list_data_sources`.\n2. Present the available source names.\n3. Ask the user to select one.\n\nIf the user names a source, call `list_data_sources` when needed to verify the current connector slug.\n\nNever claim that a source is supported unless it is returned by `list_data_sources`.\n\n### 2. Select the connected account\n\nCall `get_connector_accounts` for the selected connector.\n\nIf more than one account is available:\n\n1. Present the account names.\n2. Ask the user which account to analyze.\n3. Do not choose an account automatically.\n\nIf only one account is available, use it.\n\nUse the `external_account_id` internally for tool calls. Do not display it unless required for troubleshooting.\n\nDo not display authentication-owner names, OAuth scopes, token-expiry values, timestamps, or other internal connection metadata.\n\n### 3. Resolve the date range\n\nIf the user gives explicit dates, use those dates.\n\nIf the user gives a supported relative period, call `get_dates` with the matching preset.\n\nIf the user gives no date range, call:\n\n`get_dates` with `preset: \"last_30_days\"`\n\nUse the returned `start_date` and `end_date`.\n\nState the resolved date range in the final answer.\n\n### 4. Discover valid fields\n\nCall `list_fields` for the selected connector and account before calling `query_data`.\n\nSelect only field IDs returned by `list_fields`.\n\nIf the user names specific metrics or dimensions, use them when available.\n\nIf the user asks for a general analysis without specifying metrics, choose a balanced set of available fields appropriate for that source.\n\nExamples, only when returned by `list_fields`:\n\n- Advertising: impressions, clicks or link clicks, spend, CTR, CPC, leads, purchases, revenue, campaign name.\n- Search performance: clicks, impressions, CTR, average position, query, page.\n- Social insights: reach, impressions, engagement, followers, media or post.\n- Ecommerce: revenue, orders, average order value, products, customers.\n- Local performance: views, calls, direction requests, website actions.\n- Website performance: performance score and Core Web Vitals.\n\nDo not invent field IDs or assume that every connector supports the same metrics.\n\n### 5. Query the selected source\n\nCall `query_data` using:\n\n- The verified connector slug.\n- The selected account's `external_account_id`.\n- Valid fields returned by `list_fields`.\n- The resolved start and end dates.\n\nFor a general channel analysis, include a useful breakdown dimension when available, such as campaign, ad set, post, query, page, product, or date.\n\nDo not request more fields than needed for a clear analysis.\n\n### 6. Analyze the results\n\nEvaluate only the returned data.\n\nLook for:\n\n- Overall performance.\n- Strongest and weakest entities.\n- Meaningful differences or trends.\n- Efficiency metrics where available.\n- Missing, zero, or unusual values.\n- Opportunities supported by the data.\n\nDo not infer causation when the data only shows correlation.\n\nDo not calculate or compare currency values unless the currency context is clear.\n\n### 7. Handle missing data safely\n\nIf the query returns no rows or an error:\n\n1. Explain that no data was returned for the selected source, account, and period.\n2. Ask the user before changing the date range, account, or requested fields.\n3. Do not silently widen the period or substitute another account.\n\n### 8. Present the result\n\nBy default, provide:\n\n1. A compact performance table.\n2. A short summary of the main findings.\n3. Practical recommendations based only on the returned data.\n\nA recommended structure is:\n\n## Channel overview\nA short summary of the selected source, account, and date range.\n\n## Performance table\nUse clear metric columns and an appropriate breakdown.\n\n## Key findings\nHighlight the most important results, trends, strengths, and weaknesses.\n\n## Recommendations\nGive practical, read-only recommendations supported by the data.\n\nIf the user asks for more detail, expand the response with:\n\n- Executive summary\n- Overall performance\n- Performance by campaign, content, product, query, page, or other available dimension\n- Trends\n- Strongest and weakest results\n- Data limitations\n- Detailed recommendations\n\n### 9. Recommendations\n\nRecommendations must:\n\n- Be directly supported by the returned data.\n- Explain the observation behind each recommendation.\n- Remain advisory and read-only.\n- Never claim that Data Bloo changed a campaign, account, budget, dashboard, or source setting.\n\nUse cautious wording when the data is incomplete or a metric definition is unclear.\n\n## Additional tool rules\n\nUse `list_accounts` only when a broader overview of all connected sources is necessary.\n\nUse `get_user` only when:\n\n- The user asks about their Data Bloo plan, MCP access, usage limits, or remaining calls.\n- A tool response indicates that access or usage limits may prevent the requested analysis.\n\nDo not use `get_user` as part of a normal channel analysis.\n\n## Output quality rules\n\n- Analyze only one connector at a time.\n- Use plain language.\n- Keep the default response concise.\n- Clearly name the source, account, and date range.\n- Separate facts from recommendations.\n- Mention missing or unavailable metrics.\n- Do not expose unnecessary internal identifiers or authentication metadata.\n- Do not promote subscriptions, pricing, upgrades, or checkout links.\n"
}