Fix agent work with sql skill

[ea-rus]

Description

Updates:

• Update instructions to force llm to use mysql quiting style (backticks for identifiers)
  • also removed mentioning postgres database in main prompt for it
• disabled sql_parser_tool. @dusvyat, do we have some benchmarks that we might use to check that this change doesn't degrade quality?
  • it adds extra step for llm before executing query (increases response time).
  • also its payload is not clear: it just parses the query and render it again. db_query also shows error in case of parsing error
• Fix error if with postgresql when column names are in mixed case (commit)
• removed unused line in prompt in main template (Here is the user's question:)
• updated error messages if a wrong database name is used (instead of Table name should contain only one part)
• accept wrong quoting for 3-item identifier (handle cases like db.schema.table)
• fix getting table info for 3-item identifier (db.schema.tablle)

Tested with queries from issues below:

--- for CONN-1442
select * from financial_agent where question = "What stocks are in the portfolio?";
select * from financial_agent where question = "What's the percentage breakdown of investments by category in the portfolio";
select * from financial_agent where question = "What category has the most exposure?";
select * from financial_agent where question = "Leadership has mandated that we limit our exposure to crypto to 10%. Is the portfolio compliant and if not, what must I do to make it compliant?";

--- for CONN-1403
select * from supplychain_agent where question ="What is the current stock level of a specific product: Radiance Ritual?";
select * from supplychain_agent where question ="Which products are low in stock and need reordering?";
select * from supplychain_agent where question ="What are the recent shipment details for a SKU2?";
select * from supplychain_agent where question ="How long does it typically take for a SKU2 to be delivered?";
select * from supplychain_agent where question ="What are the details of SKU5, such as its price and description?";
select * from supplychain_agent where question ="Can you list all the products from a cosmetics category?";
select * from supplychain_agent where question ="Who are the suppliers for SKU5?";
select * from supplychain_agent where question ="What are the contact details of supplier of SKU5?";
select * from supplychain_agent where question ="Can you provide a summary of the inventory status?";
select * from supplychain_agent where question ="What are the top-selling products?";

--- for CONN-1500 
select * from my_sf_agent where question = 'show me the top 10 customers by total order value'
select * from my_sf_agent where question = 'which suppliers provide parts to a specific region'
select * from my_sf_agent where question = 'what the most frequient ordered parts'
select * from my_sf_agent where question = 'what are the columns of table customers?'

--- for CONN-1459 
select * from my_agent where question = 'does O_ORDERPRIORITY column has a null values?'

Fixes https://linear.app/mindsdb/issue/CONN-1442/bug-agentsintermittent-issues-in-executing-sql-query (#11457)

• updated instructions to use mysql dialect & fix getting sample rows

Fixes https://linear.app/mindsdb/issue/CONN-1403/agents-unable-to-retrieve-data-from-certain-tables-it-has-access-to

• fixed error message in case of wrong database

Fixes https://linear.app/mindsdb/issue/CONN-1500/bug-table-not-found-for-snowflake-tables (#11529)
Fixes https://linear.app/mindsdb/issue/CONN-1459/bug-frequently-getting-table-tbl-not-found (#11394)

• fixes for 3-item identifier

Type of change

• 🐛 Bug fix (non-breaking change which fixes an issue)

Verification Process

To ensure the changes are working as expected:

• Test Location: Specify the URL or path for testing.
• Verification Steps: Outline the steps or queries needed to validate the change. Include any data, configurations, or actions required to reproduce or see the new functionality.

Additional Media:

• I have attached a brief loom video or screenshots showcasing the new functionality or change.

Checklist:

• My code follows the style guidelines(PEP 8) of MindsDB.
• I have appropriately commented on my code, especially in complex areas.
• Necessary documentation updates are either made or tracked in issues.
• Relevant unit and integration tests are updated or added.

[entelligence-ai-pr-reviews]

🔒 Entelligence AI Vulnerability Scanner

✅ No security vulnerabilities found!

Your code passed our comprehensive security analysis.

📊 Files Analyzed: 3 files

---

[entelligence-ai-pr-reviews]

Review Summary

🏷️ Draft Comments (2)

Skipped posting 2 draft comments that were valid but scored below your review threshold (>=13/15). Feel free to update them here.

mindsdb/interfaces/skills/custom/text2sql/mindsdb_sql_toolkit.py (1)

21-102: get_tools constructs large multiline string SQL instructions on every call, causing repeated CPU and memory overhead for each agent/toolkit instantiation.

📊 Impact Scores:

• Production Impact: 3/5
• Fix Specificity: 4/5
• Urgency Impact: 2/5
• Total Score: 9/15

🤖 AI Agent Prompt (Copy & Paste Ready):

Refactor mindsdb/interfaces/skills/custom/text2sql/mindsdb_sql_toolkit.py lines 21-102: Move the large multiline SQL instruction string templates (especially for query_sql_database_tool_description) to class-level constants or static variables, and use .format() to inject dynamic values as needed. This avoids reconstructing large strings on every get_tools() call, reducing repeated CPU and memory overhead for each agent/toolkit instantiation. Preserve all logic and formatting.

---

mindsdb/interfaces/skills/sql_agent.py (1)

588-598: _get_sample_rows now uses select * instead of selecting only the provided fields, which can cause column order mismatches and incorrect CSV headers if the table has more columns than fields.

📊 Impact Scores:

• Production Impact: 4/5
• Fix Specificity: 5/5
• Urgency Impact: 3/5
• Total Score: 12/15

🤖 AI Agent Prompt (Copy & Paste Ready):

In mindsdb/interfaces/skills/sql_agent.py, lines 588-598, the `_get_sample_rows` method was changed to use `select *` instead of selecting only the provided `fields`. This can cause the returned sample rows to have columns in a different order or extra columns, leading to mismatches with the CSV header and potentially incorrect data shown to users. Please change the SQL command back to `select {', '.join(fields)} from {table} limit {self._sample_rows_in_table_info};` to ensure the sample rows match the expected fields.

---

[ea-rus]

@StpMax, @dusvyat, the PR is ready for review, I'm not going to add anything to it

The base branch was changed.