MCP server for USA National Phenology Network (NPN) API providing tools for querying phenological observation data, generating phenometrics, and accessing reference materials
Server has 10 tools with complete input schemas and descriptions, but quality is inconsistent. Tool names lack action verbs (e.g., 'status-intensity', 'magnitude-phenometrics' are noun-heavy; should be 'query-status-intensity'). Descriptions are present but generic (avg ~80 chars, below 194-char baseline). Parameter descriptions are adequate but lack format constraints, ranges, and examples of valid values. Output schemas are not documented, LLMs cannot predict response structure. Error handling is absent from visible code. One tool (export-raw-data) is WRITE-capable but lacks confirmation/dry-run pattern. SQL injection risk in query-reference-material and query-literature tools (raw SQL params). No tool annotations (readOnlyHint/destructiveHint) despite MCP support.
Tool for exporting raw cached data to file
Tool for retrieving raw cached data from previous queries
Tool for querying individual-level phenometrics data
Tool for querying magnitude phenometrics data
Tool for generating maps of phenological observation data
Tool for retrieving comments associated with observations from the NPN API
Tool for querying literature and analysis data from the NPN database
Tool names lack action verbs. 'status-intensity', 'magnitude-phenometrics', 'site-phenometrics' are noun-heavy descriptors, not verb-noun actions. LLMs infer intent from names, these names obscure what the tool does. Rename to 'query-status-intensity', 'query-magnitude-phenometrics', etc.
Output schemas not documented. Tool descriptions state what they query but do not specify response structure (fields, types, pagination). LLMs cannot plan downstream calls or extract data without knowing response shape. Add 'Returns: {field_name: type, ...}' to each tool description.
SQL injection risk in query-reference-material and query-literature. Both accept raw 'sql_query' parameter with no validation or sanitization. LLMs can be tricked into passing malicious SQL. Implement parameterized queries or a safe query builder; never expose raw SQL to agents.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | <=2025-11-25 | v2 |
Tool for querying reference materials from the NPN database
Tool for querying site-level phenometrics data
Tool for querying intensity and status data from the NPN API
export-raw-data is WRITE-capable (file export) but lacks confirmation/dry-run pattern. Agents can overwrite files without user approval. Add a 'dry_run' parameter or require explicit confirmation before writing.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite MCP support. Clients cannot distinguish safe tools from destructive ones. Add readOnlyHint=true to all query tools; add destructiveHint=true to export-raw-data.