Six tools with clear, domain-specific names (net_*) and solid descriptions (avg 150 chars). All tools have input schemas with type definitions. However, parameter descriptions are sparse or missing entirely in several tools (net_github_status, net_proxy_status, net_hosts_check, net_doctor have empty input objects with no params to describe). Output schemas are not documented, LLMs cannot predict response structure. Error handling uses a generic 'netassist error:' wrapper without recovery guidance. No tool annotations (readOnlyHint, etc.). Naming follows verb_noun pattern well; composition is sound (each tool does one thing). Baseline: avg tool description ~150 chars (within 10-1024 range), but lack of output schema documentation and sparse parameter descriptions pull the score into fair territory.

• Document output schemas for all tools. Define the structure of response fields (e.g., net_github_status returns {reachable: bool, dns: string[], tcp443: string, httpStatus: number, tcpMs?: number}). This enables LLMs to plan downstream reasoning.
• Add tool annotations: mark all six tools with readOnlyHint=true since they perform no mutations. This signals to agents that these calls are safe to retry and have no side effects.
• Enhance error handling with categorization. Instead of 'netassist error: <msg>', return {error: true, category: 'retryable'|'user_fixable'|'fatal', message: '...', suggestion: '...'}. Example: 'DNS resolution failed for github.com. Suggestion: check network connectivity or try again in 30s.'
• For net_doctor, document the structure of the report object (findings array with {level, title, detail}, suggestions array). Currently the rendering logic is in the handler; the schema is implicit.
• Add parameter constraints to net_proxy_probe: document that ports must be 1-65535 and recommend common proxy ports (10808, 10809, 7890, 7897, 8888, 1080) in the description, not just the default list.