Credentials exposed as tool parameters: refresh_token, access_token, customer_id passed as parameters in all 4 tools instead of server-side secret injection. Secrets will appear in agent logs and prompt history.

Output schemas not documented. Tools return free-form objects without schema definition, forcing LLMs to parse unstructured data. Downstream tool chains cannot reliably extract required fields (customer_id, ad_group_id, etc).

Date range parameters lack format constraints. start_date and end_date described only as 'YYYY-MM-DD format' in text, no minLength, maxLength, pattern, or format JSON Schema properties. LLMs cannot validate without formal constraints.

Duplicate/conflicting customer_id parameters. get_ad_details has both 'customer_id' in user_credentials AND 'customer_id' as a top-level parameter with unclear priority ('Top-level customer_id (preferred)'). This ambiguity forces LLMs to reason about which to use.

No error guidance for common failures. Adapters throw generic 'Customer not found or no access' and 'Ad Group {id} not found or no access' without recovery suggestions. LLMs cannot decide whether to retry, ask the user, or try a different approach.

Parameter relationships undocumented. get_account_summary's comparison_date_range is optional but depends on date_range being set first, no documentation of this dependency. get_ad_details requires both ad_group_id and ad_id, but no statement that both are required together.

Required parameter inconsistency. get_account_summary marks date_range as an object with required array ['start_date', 'end_date'], but comparison_date_range has no required array, making it unclear if both start_date and end_date are mandatory or independent.

Tool descriptions lack usage context. 'Get detailed configuration...' does not explain WHEN to call vs similar tools, or what downstream actions this enables. No mention of prerequisites (is customer authenticated? is account accessible?).