Validate classification banner markings (CUI/CONFIDENTIAL/SECRET) in documents per portion-marking rules.
Single tool 'classguard_scan' has a basic description (89 chars) and minimal schema. Input schema present but lacks detail: only 'target' (string) with generic description. No output schema documented. No parameter constraints (enum, pattern, min/max). No error handling guidance. No tool annotations. Naming is acceptable (verb_noun form) but the tool lacks the depth expected for production use. Descriptions fall short of the 194-char baseline for A+ tools.
Validate classification banner markings (CUI/CONFIDENTIAL/SECRET) in documents per portion-marking rules.. Returns JSON findings.
No output schema documented. LLM cannot predict return structure (fields, types, nesting). Breaks downstream tool chaining and forces LLM to parse unstructured JSON.
Parameter 'target' lacks constraints. No indication whether it accepts files, directories, or both; no path format guidance; no validation rules. LLM may pass invalid paths.
Tool description (89 chars) is below the 194-char baseline for A+ tools. Missing WHEN to use, WHAT it returns, and prerequisites. Insufficient for LLM selection logic.
No error handling guidance. If scan fails (permission denied, invalid path, I/O error), LLM receives no recovery hint. No categorization of retryable vs fatal errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLM cannot infer safety properties. Scan is read-only but this is not declared in schema.