Read-only Robinhood Chain market intelligence tools for AI agents.
Five tools with complete Zod schemas, clear descriptions (100-200 chars), and proper parameter validation. All tools are read-only with correct annotations. Naming follows verb_noun pattern (inspect_token, simulate_exit, compare_exit_sizes, chain_health, explain_exit_signal). Schemas use regex patterns for EVM addresses and numeric bounds. Error handling wraps all tools with try-catch and returns actionable error messages. Output structure is well-defined with compactReport() and explainReport() helpers. Minor gaps: parameter descriptions could be more explicit about constraints (e.g., 'minimum 0' not stated in description text, only in schema); no pagination guidance for future scaling; output schema documentation is implicit rather than explicit in tool registration.
Read chain ID, latest block and RPC latency without connecting a wallet.
Compare 1 to 8 USD exit sizes against the same observed primary pool.
Turn the deterministic DEEP, THIN or CRITICAL signal into evidence and next checks an AI agent can explain.
Read current pools, liquidity, volume, price and a small-position exit grade.
Estimate exit pressure for a USD position using current Robinhood Chain market data. This is a proxy, not an executable quote.
Parameter descriptions lack explicit constraint documentation. Zod schema enforces bounds (e.g., sizeUsd max 10M), but descriptions do not state 'USD position size (0 - 10,000,000)'. LLMs cannot read JSON Schema, they rely on description text.
Output schema is implicit. Tools return structured data via compactReport() and explainReport(), but the tool registration does not document the return type schema. LLMs cannot plan downstream operations without knowing what fields to expect.
compare_exit_sizes accepts 1 - 8 sizes but does not explain why the upper bound exists or what happens if the agent needs to compare more. Undocumented limits confuse LLMs.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2025-06-18+ | v2 |
Error messages are generic ('TRENCH error: ...') and do not guide recovery. Per pattern:recovery-guide, errors should suggest next steps (e.g., 'Invalid token address. Verify the 0x-prefixed EVM address and try again.').
No dry-run or confirmation pattern for tools that depend on external RPC state. If chain_health or token analysis fails due to transient RPC issues, error guidance should suggest retry with backoff.