Five rigorous reasoning protocols - debate, red team, argument audit, threat model, study sanity - free MCP plugin for Claude and ChatGPT.
All 5 tools lack formal input schemas in the source code. While mcp.json lists tool names and descriptions are present in the Protocol objects, there are NO visible JSON Schema definitions with type constraints, required fields, or parameter validation. The descriptions are lengthy (200-400+ chars) but read as protocol instructions rather than LLM-optimized tool docstrings. No output schemas documented. Tools are inferred from Protocol constants rather than explicit MCP tool registration with schemas. This violates the core requirement that every tool must have a complete, machine-readable input schema.
Audit the reasoning quality of a text. Reconstruct the argument, audit the reasoning for fallacies and biases, audit the rhetoric, and name what evidence is conspicuously absent.
Methodology and statistics sanity check. Audit a study, experiment, or statistical claim for methodological soundness, statistical validity, and hidden assumptions.
Run an adversarial expert debate on a motion. Convene 3-6 experts derived from the motion, run openings/rebuttals/evidence rounds, and deliver a verdict with the strongest dissent preserved.
Adversarial stress-test of a plan, design, strategy, product, or claim. Steelman the idea, map the attack surface, run a pre-mortem, and rank risks with mitigations or falsifying tests.
Defensive STRIDE security review. Identify threats to a system, asset, or design using the STRIDE framework (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege).
No input schemas visible in source code. All 5 tools declare only a single string parameter (motion, idea, text, system, study) with minimal type info. JSON Schema with type, required, and description fields is absent.
Tool definitions inferred from Protocol constants, not explicit MCP tool registration. No visible McpServer.tool() calls or schema registration in src/index.ts.
Descriptions are protocol instructions (200-400+ chars), not LLM-optimized tool docstrings. They explain HOW to run the protocol, not WHAT the tool does or WHEN to call it. Baseline for A+ tools: 50-200 chars, action-focused. These read as user guides, not tool descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
No output schemas documented. Tools return plain text protocol instructions; no structured response format, field types, or pagination guidance. LLMs cannot plan downstream calls or extract data reliably.
No error handling guidance. Tools return plain text; no error classification (retryable, user-fixable, fatal), no recovery hints, no actionable error messages. Agents cannot distinguish success from failure or know what to do next.