ClineMCP has 7 tools with complete input schemas and descriptions, but quality is inconsistent. Tool names follow verb_noun convention (cline_start, cline_status, cline_cancel, cline_output, cline_tail, ensure_clinerules). Descriptions are present but brief (avg ~60 chars, below the 194-char baseline). Parameters have types and descriptions, but lack constraints (enums, ranges, patterns). Output schemas are not documented, responses are JSON strings without declared structure. Error handling is minimal; most tools return JSON with error fields but lack recovery guidance. Security concern: session_id is passed as a parameter rather than extracted from request context, and no permission checks are visible. The tool composition is reasonable (single responsibility), but parameter naming could be more explicit (e.g., 'lines' should specify range 1-1000).

• Document output schemas for all tools. Define the structure of each JSON response (e.g., cline_start returns {session_id: string, status: string, started_at: string, error: string|null}).
• Add parameter constraints: 'lines' in cline_tail should be 1 - 1000 (default 20); 'step_id' should be a positive integer; 'status' values should be an enum (running, completed, failed, cancelled).
• Expand descriptions to 100 - 200 chars. Example: 'cline_cancel: Terminate the active Cline subprocess for a session. Use this to stop a long-running task. Returns success status and whether the session was running.'
• Add recovery guidance to error responses. E.g., 'Session not found: {session_id}. Call cline_start() to create a new session, or check the session_id spelling.'
• Implement permission checks: verify the calling agent/user has authority to access/modify each session. Log all access attempts for audit trails.

Inferred effective spec: <=2025-11-25.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained