Deterministic MCP policy and JSON-RPC transcript simulator with tamper-evident audits.
Two tools with minimal documentation. Both have descriptions (good), but input schemas lack parameter-level descriptions entirely. trustline.simulate accepts 'policy' and 'transcript' objects with only type and high-level descriptions; trustline.verify accepts 'audit' with minimal context. No output schemas documented. Parameter descriptions do not explain format, constraints, or expected structure. Naming is domain-specific but not action-verb-based (simulate/verify are verbs, but lack clarity on what data structures they operate on). No error handling guidance, no examples of valid inputs, no recovery paths.
Simulates a JSON-RPC transcript against a Trustline policy and produces a deterministic audit trail with tamper-evident hashing.
Verifies the integrity and validity of a Trustline audit bundle by checking the cryptographic hash chain and policy digest.
Input parameters lack descriptions. 'policy' and 'transcript' in trustline.simulate have only type and high-level schema notes; no guidance on format, constraints, or valid values. LLMs cannot infer what constitutes a valid policy object or transcript array.
No output schemas documented. Neither tool specifies what fields the response contains, their types, or structure. Agents cannot plan downstream operations or extract required data.
Tool descriptions lack actionable context. 'Simulates a JSON-RPC transcript...' does not explain when to call this tool, what prerequisites exist, or what the audit trail contains. Descriptions should answer: What does it do? When should I call it? What do I get back?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
No error handling guidance. No documentation of what errors can occur, how to recover, or what invalid inputs look like. An LLM receiving a validation error has no path forward.
Naming does not follow verb_noun convention. 'trustline.simulate' and 'trustline.verify' are verbs but lack object clarity. Better: 'simulate_trustline_policy' or 'verify_audit_bundle' to make the subject explicit.