The server defines 3 tools with complete schemas and reasonable descriptions, but has multiple gaps in parameter descriptions, lacks error handling guidance, and missing documentation for output structures. Naming follows verb_noun conventions. Descriptions are adequate (100-300 chars) but could be more precise about LLM selection criteria. All tools are READ_ONLY, lowering security risk. Parameter descriptions exist but are sometimes vague ('Путь к директории или файлу' lacks format constraints). Output structures are returned as formatted strings rather than structured JSON, reducing downstream composability.

• Refactor all three tools to return structured JSON objects with typed fields instead of formatted strings. E.g., scan_tech_debt should return {results: [{file, line, marker, priority, severity, text, context}, ...], total_found, files_scanned, stats: {high: N, medium: N, low: N}} instead of a markdown string.
• Add formal enum constraints to schema for scan_tech_debt.priority_filter and audit_dockerfile.strict. Move constraint documentation from free-text descriptions into JSON Schema enum arrays.
• Clarify parameter formats: scan_tech_debt.extensions should state 'Comma-separated, no spaces (e.g., .py,.js,.ts)'. generate_release_notes.from_ref and to_ref should state 'Git ref: branch, tag, or commit hash (40-char SHA1 or short form)'.
• Add per-tool error handling function with recovery hints. Instead of 'Path not found: {path}', return {error: 'PATH_NOT_FOUND', message: 'Directory does not exist', suggestion: 'Check path format; use absolute paths or relative to current directory', recoverable: true}.
• Document output schema for each tool in the tool description. E.g., 'Returns JSON object with fields: results (array of {file, line, marker, priority, text}), stats (object with counts), total_found (integer).'
• Add per-item IDs to results arrays to support future tool chaining. E.g., scan_tech_debt results should include issue_id field for potential 'resolve_tech_debt(issue_id)' tool.
• Add selection criteria to tool descriptions. E.g., 'Use scan_tech_debt to find code markers (TODO/FIXME). Use audit_dockerfile to check container best practices. Use generate_release_notes to summarize git history for users.'