Three tools with mixed quality. bazaar_search and bazaar_pay_and_call have detailed schemas and descriptions (150-200 chars), but summarize_text is trivial (11 chars). All tools have proper input schemas with types and descriptions. bazaar_pay_and_call uses nested object schemas (expected.properties) which is correct but complex. No error handling guidance, no output schemas documented, and no tool annotations (readOnlyHint/destructiveHint). bazaar_pay_and_call is a WRITE operation that modifies blockchain state but lacks confirmation/dry-run pattern.

• Expand summarize_text description to 50-150 chars explaining when to use it (e.g., 'Condense a long text block into a single-line summary for display or logging. Use when you need a brief overview without full details.').
• Document output schemas for all three tools. Example for bazaar_search: 'Returns {results: [{resource: string, toolName: string, paymentTerms: {network, scheme, asset, amount}, inputSchema: object}], nextCursor?: string, abstention?: {reason: string}}'.
• Add tool annotations to input schemas: bazaar_search with readOnlyHint=true, bazaar_pay_and_call with destructiveHint=true. This signals to agents which tools are safe to retry.
• Add error handling guidance to bazaar_pay_and_call: 'If live price exceeds maxAmount, returns error with fields {reason: string, livePrice: string, maxAmount: string}. Agent should ask user before retrying with higher maxAmount.'
• Implement dry-run or confirmation pattern for bazaar_pay_and_call: add optional 'dryRun: boolean' parameter. When true, returns payment terms and required signature without executing the transaction.