Borrowing it
Nothing to install: this file belongs to ProfSynapse/nexus. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ProfSynapse/nexus/main/.claude/skills/nexus-model-eval/SKILL.mdgit clone --depth 1 https://github.com/ProfSynapse/nexusWrote 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.
[](https://agentmods.dev/skills/profsynapse/nexus/nexus-model-eval)<a href="https://agentmods.dev/skills/profsynapse/nexus/nexus-model-eval"><img src="https://agentmods.dev/badge/skills/profsynapse/nexus/nexus-model-eval/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.
<a href="https://agentmods.dev/skills/profsynapse/nexus/nexus-model-eval"><img src="https://agentmods.dev/badge/skills/profsynapse/nexus/nexus-model-eval.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00070 | $0.00986 |
| Opus 5.5 | $0.00028 | $0.00394 |
| Sonnet 5.5 | $0.00014 | $0.00197 |
| Haiku 4.5 | $0.00007 | $0.00099 |
Grade A, and why
nexus-model-eval 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.
How it starts
The opening of the file, as written. The whole thing — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Nexus model eval
Context: the harness in tests/eval/ shows a model the same two tools the app
does — getTools for discovery, useTools for execution — and grades the calls
it makes, not the prose it writes. This skill owns the verdict: which models to
run, and what a FAIL actually means. Running, configuring and extending the
harness itself belongs to nexus-eval-harness. This file routes; detail loads
when you take the path.
Workflow
- Get current truth before running anything. A model cannot be graded on a
fixture no model can satisfy, and the fixture set moves:
A non-zero exit from the scenario checker means some scenario can never pass; resolve that first, and the fix belongs tols tests/eval/scenarios/ tests/eval/configs/ python3 .claude/skills/nexus-eval-harness/scripts/check_scenarios.py python3 .claude/skills/nexus-model-eval/scripts/check_advertised_tools.pynexus-eval-harness, not to this run. The advertised-tools gap is not a defect — it is the list of correct model behaviors this harness punishes, and you will need it in step 3. - Run the grade:
protocols/grade-models.md. Read it before you start; a summarized procedure is one you will improvise, and every scenario in the matrix costs live, billed API calls. - You MUST attribute every failure before you report a number:
protocols/attribute-failures.md. The harness fails models for things the model did not do, so a raw pass rate with unread failures is not a grade.scripts/summarize_eval.py --labelsrefuses to sign off while any failure is unlabelled. - Report both numbers — raw pass rate and the attributed rate that charges only
model-failureverdicts — plus what the excluded failures actually were. One number alone is either unfair to the model or unfair to the reader. - At the end of a session that used this skill, run
protocols/self-refine.md.
Map
protocols/the procedures:grade-models.md(target list → run → artifacts),attribute-failures.md(FAIL → verdict → defensible grade),self-refine.md.references/read on demand:what-is-graded.md(what makes a scenario pass, what a "turn" counts, how retries and exclusions move the number),harness-artifacts.md(symptom → cause → proof for failures the model did not cause — read this before blaming any model).scripts/run them, do not reimplement:scripts/check_advertised_tools.py— the commands the eval system prompt tells the model to use that the executor cannot run, so obeying the prompt scores as a hallucination.scripts/preflight_models.py— do these slugs exist, before the run spends money proving they do not.scripts/summarize_eval.py— report JSON → per-model rollup, bucketed failures, and an attribution that is checked rather than asserted.
refinement-log.mdwhat past sessions changed here and why.
What ships with it
9 files 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.
- protocols/attribute-failures.md 3.5 KB
- protocols/grade-models.md 4.2 KB
- protocols/self-refine.md 2.3 KB
- references/harness-artifacts.md 5.0 KB
- references/what-is-graded.md 4.7 KB
- refinement-log.md 1.4 KB
- scripts/check_advertised_tools.py 8.6 KB runs code
- scripts/preflight_models.py 4.4 KB runs code
- scripts/summarize_eval.py 10 KB runs code
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 · 69 lines · 70 tokens per session scan A 3afecc81af22
nexus-model-eval is a skill published in the GitHub repository ProfSynapse/nexus (158 stars, last pushed 5d ago), licensed MIT. It adds 70 tokens to every session and 986 once invoked, about $0.0003 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.
Other skills, from other repositories
cli-eval
Create and run evaluation suites, watch live benchmark progress, view scorecards, compare model performance, and integrate eval runs with CI workflows from the CLI.
emerging-techniques-model-merging
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task…
darwinian-evolver
Evolve prompts/regex/SQL/code with Imbue's evolution loop.
dummy-dataset
Generate realistic dummy datasets for testing with customizable columns, constraints, and output formats (CSV, JSON, SQL, Python script). Use when creating test data, building mock datasets, or generating sample data for development and demos.
validate
Validate Semantica pipelines, extraction quality, graph schemas, and ontology consistency. Returns structured error/warning checklists. Uses PipelineValidator, PipelineBuilder.validatepipeline(), GraphValidator, and OntologyValidator. Sub-commands: pipeline, step, dependencies, extraction, graph, ontology, performance.
agent-evals
Build automated evaluation suites for AI agents using golden datasets, rubrics, and regression gates. Use when shipping agent features, validating prompt changes, or gating deployments on quality.