DeltaLens provides 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (build_graph, get_delta_context, classify_change). However, descriptions are terse (avg ~80 chars, below the 194-char baseline), lack WHEN-to-use guidance, and omit output schema documentation. Parameters are typed but lack detailed constraints (e.g., token_budget has no min/max bounds). Error handling and recovery guidance are absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. The server is functional but lacks LLM-optimization depth expected of production tools.

• Add output schema documentation to all tools. E.g., get_delta_context should document: {context: string, tier: 'critical'|'important'|'supporting', tokens_used: number, nodes_included: string[]}.
• Expand descriptions to 100 - 200 chars with WHEN-to-use guidance. E.g., 'Get scored, tiered context for changed files. Call this after classify_change to prioritize which files to review first. Returns minimum viable context within token budget.'
• Add min/max bounds to numeric parameters: token_budget (1 - 100000), top_k (1 - 100), page_size (1 - 50). Include in parameter description: 'Maximum token budget (1 - 100000, default 8000)'.
• Annotate tools with readOnlyHint and destructiveHint. build_graph should have destructiveHint=true (overwrites graph); all others readOnlyHint=true. This guides agent retry logic and safety reasoning.
• Add error recovery guidance to build_graph and search_nodes. E.g., 'If parsing fails on large repos, try with ignore_patterns to exclude vendor/ or node_modules/. If search times out, reduce top_k or use_embeddings=false.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry