Loading ToolBench
Interact with Veritas Acta — contribute, challenge, discover, verify, and query contestable public records.
Server has 6 tools with explicit Zod schemas and descriptions. Naming follows verb_noun pattern (acta_contribute, acta_respond, acta_query, acta_discover, acta_verify, acta_export). Descriptions are present and substantive (100-200 chars typical). However, 4 of 6 tools (acta_query, acta_discover, acta_verify, acta_export) have minimal descriptions with no visible input schema in the provided source code, only tool names and brief descriptions are shown. The two write tools (acta_contribute, acta_respond) have complete Zod schemas with well-described parameters. Error handling returns structured feedback but lacks recovery guidance (e.g., 'try search_users() first'). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Output is formatted as markdown tables, not structured JSON.
Submit a typed contribution to the Acta public record. Questions have no evidence burden. Factual claims require a source or reasoning. Opinion/hypothesis claims require uncertainty. Predictions require resolution criteria, date, source, and rule.
Find actionable work: contested records, claims needing evidence, etc.
Export full chain for independent verification
Browse/search topics, entries, single entry lookup
Submit a response to an existing entry. Evidence requires source + stance. Challenges require target_assertion + basis + argument (min 20 chars). Updates require update_type. Resolutions require outcome + source + resolution_type.
Verify hash chain integrity for a topic
Four read-only tools (acta_query, acta_discover, acta_verify, acta_export) have no visible input schemas in source code. Only tool names and brief descriptions provided. Cannot verify parameter types, constraints, or required fields.
Descriptions for acta_query, acta_discover, acta_verify, acta_export are under 20 characters ('Browse/search topics...', 'Find actionable work...', 'Verify hash chain...', 'Export full chain...'). These lack context for LLM tool selection.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. acta_contribute and acta_respond are destructive write operations but lack destructiveHint annotation. acta_query, acta_discover, acta_verify, acta_export are read-only but lack readOnlyHint.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Error responses return markdown text, not structured JSON. acta_contribute returns 'Contribution failed (HTTP 400): ...' as plain text. LLMs cannot reliably parse unstructured error text for recovery logic.
No recovery guidance in error messages. When a contribution fails, the error does not suggest next steps (e.g., 'Check topic slug format' or 'Verify source URL is valid'). Per pattern:recovery-guide, errors should guide the LLM toward resolution.