Cache App exposes 7 tools with mixed quality. Tool names follow verb_noun convention (get*, list*, search*, update*), which is good. However, descriptions vary widely in depth and actionability. Most tools have basic descriptions (50-150 chars), but lack detail on when to use them vs. alternatives, prerequisites, or expected output structure. Input schemas are present and typed, but several parameters lack descriptions or constraints. The update_composer tool is particularly vague about what patch structure is valid. Error handling and output schemas are not documented. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels.

• Document output schema for each tool. Example: github_repo should return {stars: number, forks: number, open_issues: number, language: string, description: string}. Include field types and whether fields are always present.
• Expand update_composer patch parameter: define the valid patch structure (e.g., JSON Schema for allowed keys, types, and constraints). Provide 1-2 concrete examples of valid patches.
• Add error handling guidance to web_fetch and web_search: 'If the URL is blocked by SSRF policy, try a different domain. If rate-limited, wait 60s before retrying.'
• Add tool annotations: mark update_composer with destructiveHint=true; mark all read-only tools with readOnlyHint=true. This enables agents to reason about safety.
• Clarify pagination for listAutomationPayloadItems and search_library: document what 'truncated' means, whether cursor is opaque, and the maximum limit value.