Two tools with complete input schemas and descriptions, but significant gaps in output documentation and error handling. preflight_workspace has a well-structured schema with 5 parameters, all typed and described; record_run has 8 parameters with enums and clear intent. However, neither tool documents return schemas, error cases, or recovery guidance. Naming is clear and verb-forward (preflight_, record_). Descriptions are adequate (80-120 chars) but lack WHEN/WHY context. No security annotations, no idempotency hints, no pagination guidance. Both tools are read-only or append-only (low risk), but lack the polish expected of production tools.

• Document return schemas for both tools. preflight_workspace should specify: {root: string, limits: {max_depth, max_entries}, counts: {directories, files, symlinks, other}, bytes_seen: number, largest_files: [{path, bytes}], errors: [{path, error}], truncated: boolean}.
• Add error handling guidance: 'If root does not exist, returns error field. If max_entries exceeded, truncated=true, call again with higher limit or narrower scope.'
• Enhance preflight_workspace description: 'Call this first to audit workspace structure before deployment. Returns file counts, largest files, and excluded directories. Safe to call repeatedly.'
• Add array item constraints to record_run: 'changed, check, error, next, unverified, version each accept 0-100 string items, max 1000 chars each. Redaction applied to all values.'
• Declare tool annotations: preflight_workspace with readOnlyHint=true; record_run with idempotentHint=true (same input always appends same record).