WorkloadTruth MCP Server

Overview Schema Source Related Servers Discussions Admin

run_benchmark

Measure classifier accuracy against synthetic GPU telemetry, clean and under evasion, to sanity-check robustness before reporting accuracy or modifying rules.

Instructions

Measure the shipped rule-based classifier's accuracy against documented synthetic GPU telemetry, both clean and under a deliberate evasion transform that mimics an operator disguising a training job as inference.

Call this to report or sanity-check classifier robustness (e.g. before citing accuracy numbers, or after changing a threshold in classifier/rules.py). Do not call it to classify a live workload -- use classify_workload for that; this tool never touches real GPU telemetry.

Side effects: none. Purely computational, no files written, no network calls, no GPU access. Deterministic and idempotent -- the same trials/window arguments reproduce the same synthetic results every call. Runtime scales with trials; the defaults (50 trials, window 30) finish in a few seconds.

Parameters: trials -- trials run per profile/evasion cell. window -- telemetry samples per classification window. Example calls: {} for the documented defaults; {"trials": 200, "window": 60} for a slower, higher-confidence accuracy read.

Returns a dict with source ("synthetic"), a note warning these numbers are not comparable to any real-hardware benchmark, window_size, trials_per_cell, clean_accuracy, evasion_accuracy, and cells (a list of per-profile, per-evasion-condition {profile, evasion, trials, correct, accuracy} breakdowns).

Input Schema

TableJSON Schema

Name	Required	Description	Default
`trials`	No
`window`	No

Output Schema

TableJSON Schema

Name	Required	Description	Default
No arguments

Schema Changelog

Changes observed during successful MCP inspections.

First observedv0.2.0Sep 27, 2026

TDQS

A4.7/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations are absent, so the description carries the full burden. It discloses side effects (none: no files, network, GPU), determinism, and scaling with trials. This is strong coverage for a computational tool. Minor gap: it doesn't mention error conditions or edge cases (e.g., invalid inputs), but for this use case the disclosure is thorough.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is well-structured: opening sentence states purpose, then usage guidance, then side effects/determinism, then parameter explanation with examples, and finally the return format. Front-loaded with the core purpose, every sentence earns its place, and no redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The tool has an output schema (provided) and only two optional parameters. The description explains the return keys (with semantics of the synthetic source) and parameter effects. Even though the output schema exists, the description adds the warning about non-comparability to real hardware, which is essential context. Complete for an agent to invoke correctly.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 0%, so the description must explain parameters. It does: `trials` per profile/evasion cell and `window` per classification window, with example calls and default values. This adds meaning beyond the schema's title/defaults. A slight deduction for not specifying valid ranges or units, but the examples suffice.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool measures classifier accuracy on synthetic telemetry under clean and evasion conditions, and distinguishes it from classifying live workloads. It names the sibling `classify_workload` explicitly, so the agent can tell them apart without opening schemas.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It explicitly says when to use the tool (before citing accuracy, after changing thresholds), and when not to (for live workload classification — use `classify_workload`). It also states the deterministic/idempotent nature and runtime scaling, giving clear context for invocation.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server