The MCP server exposes a single tool 'scan' with minimal definition quality. The tool name lacks a clear action verb (should be 'scan_target' or 'crack_hash'). The input schema is present but extremely sparse: only a 'target' parameter with a generic string type and vague description ('Target path or hash to scan/crack'). No output schema is documented. The tool description is borrowed from the package metadata and does not explain what the tool actually does when invoked, when to use it, or what it returns. No parameter constraints, enums, or validation guidance. The implementation delegates to cognis_core.mcp.build_mcp_server(), which is not visible in the provided source, making it impossible to verify actual tool registration, error handling, or output structure. Per HARD SCORING RULES: tool lacks documented output schema (schema score capped at 0), description is generic and under 50 chars of actionable guidance (description score capped at 20).

• Rename tool to 'crack_hash' or 'queue_hash_for_cracking' to clarify the action and domain.
• Document the complete output schema: e.g., {"job_id": string, "status": enum[queued|running|completed], "results": [{"hash": string, "plaintext": string, "algorithm": string}], "queued_at": ISO8601, "completed_at": ISO8601}.
• Add parameter constraints: 'target' should specify accepted formats (MD5, SHA1, SHA256, file path), max length, and character restrictions. Use enum or regex pattern.
• Expand tool description to 100 - 200 chars: 'Queue a password hash or file path for cracking via hashcat. Returns a job ID for status polling. Supports MD5, SHA1, SHA256, bcrypt. Requires valid hash format or readable file path.'
• Expose error handling: document what happens if hash format is invalid, file is unreadable, or queue is full. Provide recovery guidance (e.g., 'Invalid hash format. Supported: MD5 (32 hex), SHA1 (40 hex), SHA256 (64 hex).').