Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names follow verb_noun convention (get_trending, get_reddit_intel), which is good. However, parameter descriptions are minimal or absent. The 'track' parameter in both tools lacks actionable guidance on when to use each track value. Output schemas are not documented, LLMs cannot predict what fields to expect or plan downstream operations. Error handling is absent; no guidance on retryability or recovery. Sensitive-topic filtering is mentioned in description but not exposed as a parameter, limiting agent control.

• Add explicit output schema documentation for both tools. Example for get_trending: 'Returns {date: ISO string, sources_ok: [string], 亲子教育: [string], AI工具: [string], ...}'. This enables LLMs to plan downstream operations.
• Enhance 'track' parameter description: 'Content track to filter trending topics. Use 亲子教育 for parenting/education topics (Chinese), AI工具 for AI/productivity (Chinese), EN_kids for parenting (English), EN_ai for AI/work (English), all for all tracks combined, raw for unfiltered source data.'
• Add description to 'sources' parameter: 'Data sources to query. weibo/baidu/zhihu for Chinese trends, google_trends for US trends. If omitted, queries all available sources. Returns sources_ok field indicating which succeeded.'
• Implement error classification in responses. Return {sources_ok: [...], sources_failed: {weibo: 'timeout', ...}, partial_results: true} so LLMs know what succeeded and whether to retry.
• Add optional 'sensitivity_level' parameter (enum: strict|moderate|permissive) to let agents control filtering, with description explaining what each level filters.