Single tool 'milstdlint_scan' has a basic description (67 chars) and minimal input schema. The tool name uses verb_noun pattern correctly ('scan' action), but the description lacks critical details: no mention of what 'findings' structure is returned, no guidance on when to use it, no error recovery hints. Input schema defines only 'target' (string) with a 2-word description, no type constraints, no format guidance, no examples of valid paths. Output schema is completely undocumented; the tool returns JSON but the LLM has no visibility into the structure of 'findings' or 'score' fields. No error handling guidance. No pagination, no limits on result size. Baseline: 194 chars avg description, 72 chars avg param annotation; this tool is well below both. The tool is read-only (safe), but lacks the depth expected for production use.

• Expand tool description to 100 - 150 chars: 'Scan documents or directories for MIL-STD / DoD compliance violations. Returns structured findings with rule ID, severity (high/medium/low), and file location. Use this to audit document formatting and classification markings before submission.'
• Document output schema in tool docstring: 'Returns {"tool": "milstdlint", "findings": [{"id": "GEN-001", "sev": "high", "where": "/path/to/file"}], "score": <count>}'. Include field descriptions so LLM knows what each field means.
• Add format guidance to 'target' parameter: 'Path to document or directory (absolute or relative). Accepts .txt, .pdf, .docx files and directories (scanned recursively). Example: "/docs/project" or "./compliance-audit.txt".'
• Add error handling guidance: 'If path is invalid or inaccessible, returns error with available alternatives. If no findings, returns empty findings array with score=0. Retryable on timeout; user-fixable on permission errors.'
• Add result limit: 'Returns up to 1000 findings per scan. For large codebases, scan subdirectories separately or filter by rule ID to manage result size.'