Top-N SHAP feature attributions for the ML ensemble score on a ticker: drivers[] ranked by |SHAP| with feature (e.g. beta_cma, vol_21d), signed shap_value, direction (bullish/bearish/neutral). PRIMARY tool for 'why is the model bullish/bearish on X?'. Nightly run; default top=5, max 20; available:false outside the ML universe. 5min cache.
| Name | Required | Description | Default |
|---|---|---|---|
| top | No | ||
| ticker | Yes | Path parameter 'ticker' (required). |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden of behavioral disclosure. It discloses the ranking mechanism (by |SHAP|), output fields, default and maximum limits, update frequency (nightly), availability limitation (ML universe only), and caching (5 minute). These details go well beyond a minimal description and give the agent realistic expectations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately sized and front-loaded. The first sentence states the core purpose and output structure, followed by concise operational notes. Every phrase adds value with no redundancy or filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple read-only retrieval tool with no output schema and no annotations, the description is quite complete. It covers input (ticker), output structure (drivers[] with feature, shap_value, direction), limits, freshness, availability, and cache. The absence of explicit error handling or auth requirements is a minor gap, but not critical for this tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The description adds critical meaning to the 'top' parameter by specifying 'default top=5, max 20', which the schema omits entirely. It also confirms 'ticker' as the subject. However, the schema types 'top' as string while the description implies a numeric count, creating ambiguity. The 50% schema coverage is only partially compensated.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description opens with 'Top-N SHAP feature attributions for the ML ensemble score on a ticker', a specific verb+resource phrase. It clearly states the output structure (drivers[] ranked by |SHAP| with feature, signed shap_value, direction) and identifies itself as the 'PRIMARY tool for why is the model bullish/bearish on X?', distinguishing it from sibling ML tools.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly frames its primary use case: 'PRIMARY tool for why is the model bullish/bearish on X?'. It also provides operational constraints: 'Nightly run; default top=5, max 20; available:false outside the ML universe; 5min cache.' While no alternatives are named, the primary designation and constraints offer strong contextual usage guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
With 336 tools, there is substantial overlap. Over a dozen health/status tools share nearly identical 'is the system healthy?' descriptions (e.g., tengu_status, tengu_ready, tengu_ml_health, tengu_v3_system_health, tengu_v3_stream_status), and multiple single-ticker analysis (tengu_ml_predict, tengu_copilot_score_ticker, tengu_v3_intel_ml_prediction) and top-picks (tengu_copilot_top_picks, tengu_ml_top_picks, tengu_v3_trade_setups) tools have poorly defined boundaries. Agents would frequently misselect.
The server mixes no-version (tengu_crypto), v2 (tengu_v2_drift), v3 (tengu_v3_intel_*), and copilot (tengu_copilot_*) families, and within families there is inconsistent verb/noun ordering (tengu_v3_research_fetch_url vs tengu_v3_news_summary). While subfamilies like tengu_v3_private_markets_* are internally consistent, the overall naming pattern is chaotic and unpredictable.
336 tools is far beyond any reasonable tool set size, even for an all-in-one financial data platform. This extreme count creates choice paralysis, high latency in tool selection, and makes the server effectively unusable for autonomous agents. The calibration guideline marks 50+ as extreme; this is nearly 7x that threshold.
The platform covers a vast domain: equity and crypto prices, fundamentals, insider trading, options, news (including crypto and FX), private markets, streaming data, risk metrics, and execution planning. There are minor gaps (no direct multi-ticker comparison tool, no order placement), but the surface is remarkably comprehensive for an analysis-focused server.
