What is evaluation-benchmarking?

A WorldForge helper for running evaluations and benchmarks. Evaluations check expected behavior, while benchmarks measure supported provider operations such as prediction, embedding, scoring, and policy decisions.

Use it to change evaluation suites, benchmark harnesses, fixtures, budget checks, metric handling, and JSON, Markdown, or CSV report output.

It keeps test inputs, metrics, reports, and budget limits reproducible so results support only claims that the available evidence justifies.

How AbdelStark/worldforge configures Codex

Reuse

Borrowing it

Nothing to install: this file belongs to AbdelStark/worldforge. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/AbdelStark/worldforge/main/.codex/skills/evaluation-benchmarking/SKILL.md

Clone the repo

git clone --depth 1 https://github.com/AbdelStark/worldforge

Made for: Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/abdelstark/worldforge/evaluation-benchmarking/github.svg)](https://agentmods.dev/skills/abdelstark/worldforge/evaluation-benchmarking)

Your own site

<a href="https://agentmods.dev/skills/abdelstark/worldforge/evaluation-benchmarking"><img src="https://agentmods.dev/badge/skills/abdelstark/worldforge/evaluation-benchmarking/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/abdelstark/worldforge/evaluation-benchmarking"><img src="https://agentmods.dev/badge/skills/abdelstark/worldforge/evaluation-benchmarking.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 53 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 600 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits

NVIDIA SkillSpector pass 7 Sept 2026

How audits are shown

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00053	$0.00600
Opus 5.5	$0.00021	$0.00240
Sonnet 5.5	$0.00011	$0.00120
Haiku 4.5	$0.00005	$0.00060

Measured 2mo ago against content hash 1cbb700b7b9b, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

evaluation-benchmarking scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.codex/skills/evaluation-benchmarking/SKILL.md · 48 lines

How it starts

The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Evaluation And Benchmarking

Ground Rules

Built-in eval suites are deterministic contract harnesses, not physical-fidelity or media-quality evidence.
Benchmark direct provider operations only: predict, embed, score, and policy.
plan() is a WorldForge facade workflow. Do not route benchmark score or policy through it.
Preserve BenchmarkBudget non-zero exit behavior on violations.
Preserve claim-boundary and metric-semantics metadata in JSON, Markdown, and CSV renderers.
Treat benchmark/eval output as evidence only when the fixture, command, provider surface, and renderer semantics are all current and inspectable.

Workflow

Read src/worldforge/evaluation/suites.py for eval changes or src/worldforge/benchmark.py for benchmark changes.
Validate inputs eagerly through BenchmarkInputs and load_benchmark_inputs(...); reject unknown keys and non-finite metrics.
Keep examples/benchmark-inputs.json and examples/benchmark-budget.json reproducible and checkout-safe.
Keep score and policy payloads JSON-native; for provider-native tensors or arrays, preview type and shape rather than forcing raw encoding.
If operation surfaces or CLI text change, update help snapshots, harness diagnostics, README, docs/src/benchmarking.md, docs/src/api/python.md, docs/src/playbooks.md, and changelog together.
Test direct operation behavior, input parsing, budget pass/fail paths, and renderer output.

Definition Of Done

The changed suite or benchmark path has focused tests for success and failure behavior.
Rendered artifacts remain JSON-native and internally coherent before Markdown/CSV output.
Claims in README/docs/changelog match the actual command and provider used.
The final validation includes the focused tests plus docs/provider checks when public text changed.

Metric Semantics

Latency is process-local wall-clock timing for successful samples.
Retry counts come from emitted ProviderEvent records.
Throughput is successful samples over elapsed time.
Event rows can aggregate attempts; sum request_count when reporting actual request volume.

Read the full file on GitHub · 48 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

agents/openai.yaml 234 B

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 48 lines · 53 tokens per session scan A 1cbb700b7b9b

Subscribe to this mod's changes

evaluation-benchmarking is a skill published in the GitHub repository AbdelStark/worldforge (108 stars, last pushed 1mo ago), licensed MIT. It adds 53 tokens to every session and 600 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Analytics and metrics Testing jepa machine-learning physical-ai robotics world-models

trulens-evaluation-workflow

Systematically evaluate your LLM application with TruLens.

truera/trulens · 18 tokens

trulens-running-evaluations

Execute TruLens evaluations and view results.

truera/trulens · 15 tokens

edd

Eval-driven development loop for this repo (rag-loop, an aircraft-systems RAG app on LangGraph, OpenAI, Qdrant Cloud, and TruLens). Use when the user wants to run an experiment, inspect feedback metrics, or decide keep-or-reject on a config/prompt change.

truera/trulens · 63 tokens

worldforge: Skill for Codex