Loading ToolBench

Reading the current corpus…

elicitationtestermcp MCP server · D · ToolBench

elicitationtestermcp

github.com/mappedsky/elicitationtestermcp

A test fixture for MCP elicitation. It asks for input it does not want and never keeps a value, only its length and digest. Exposes a catalogue of MCP elicitation shapes as tools so an MCP client can be driven through every one of them on demand.

2026-07-28+STREAMABLE_HTTP7 tools

53/100

Definition Quality55

Protocol Readiness72

Supportability36

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · detected: Sampling, Roots

Deprecated protocol patterns detected

Summary

Strong foundation with clear naming, comprehensive descriptions, and proper tool annotations. All 7 tools have verb-based names (list_, run_, custom_, session_, exchange_, reset_) and detailed descriptions (100-250 chars). Input schemas are present for most tools with typed parameters. Tool annotations correctly applied (ReadOnlyHint, IdempotentHint). Key gaps: output schemas not documented in source; some parameter descriptions could be more prescriptive about constraints and error recovery; no explicit error handling guidance in descriptions.

Tools (7)

custom_formwritesource verified57/100

Send a form elicitation carrying a schema supplied verbatim, for probing a shape the catalogue does not cover. The schema is passed through byte for byte, including one that is not valid JSON Schema.

custom_urlwritesource verified57/100

Send a URL elicitation, or a -32042 URL-elicitation-required error, for an arbitrary URL.

exchange_logread onlysource verified82/100

Report what the server sent and what came back, newest first. This is the read-back when the client redacts submitted values from its own transcript.

list_scenariosread onlysource verified85/100

List the elicitation scenarios this server can run, with what each one emits and what a conforming client should do about it.

reset_logdestructivesource verified55/100

Discard the exchange log.

run_scenariowritesource verified80/100

Run one scenario from list_scenarios. Most scenarios ask for input; answering resumes the call, and the result reports the action and a fingerprint of each value rather than the value itself.

session_info

Issues (5)

HIGHtool-output-schema ↗

Output schemas not documented. Tool descriptions explain what tools do but not what they return. LLMs cannot plan downstream calls or extract fields without knowing response structure.

list_scenariosrun_scenariocustom_formcustom_urlsession_infoexchange_logreset_log

MEDIUMconstrained-input ↗

Parameter constraints not fully specified. 'scenario' param in run_scenario lacks format guidance; 'delivery' enum values (auto, deferred, direct) are mentioned but not formally constrained; 'limit' in exchange_log has no min/max bounds.

run_scenarioexchange_log

MEDIUMrecovery-guide ↗

Error handling guidance missing. Descriptions do not explain what errors are possible, when to retry, or what the LLM should do if a scenario fails. E.g., run_scenario could fail if scenario ID is invalid or delivery mode is unsupported.

run_scenariocustom_formcustom_url

Recommendations

Document output schemas for all tools. For list_scenarios, specify: returns array of {id, kind, expect, description}. For run_scenario, specify: returns {action, fingerprints, state}. For exchange_log, specify: returns {entries: [{sent, received, timestamp}], total}.
Add enum constraints to parameters. Change 'delivery' description to formally declare enum: ['auto', 'deferred', 'direct']. Add min/max to 'limit' (e.g., 1-100). Add pattern to 'scenario' (e.g., '^[a-z_]+/[a-z_]+$').
Enhance error guidance in descriptions. E.g., run_scenario: 'If scenario ID is invalid, returns error -32602. If delivery mode is unsupported by client, returns error -32042. Retry with session_info to check negotiated capabilities.'
Add input schema to reset_log. Specify that it takes no required parameters (or optional confirmation flag). Make schema visible in source code.
Implement confirmation pattern for destructive tools. Add optional 'confirm' boolean parameter to reset_log (default false). When false, return {would_delete: N_entries, confirm_with: {confirm: true}}. Only delete when confirm=true.

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Sampling (deprecated) - integrate directly with the LLM provider API

Score history

Overall score trend

First recorded score · v2 rubric

53/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	53	2026-07-28+	v2