Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add matteotitta/genesys-skills --skill eval-harnessgit clone --depth 1 https://github.com/matteotitta/genesys-skillsWrote 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/matteotitta/genesys-skills/eval-harness)<a href="https://agentmods.dev/skills/matteotitta/genesys-skills/eval-harness"><img src="https://agentmods.dev/badge/skills/matteotitta/genesys-skills/eval-harness/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/matteotitta/genesys-skills/eval-harness"><img src="https://agentmods.dev/badge/skills/matteotitta/genesys-skills/eval-harness.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 | $0.00144 | $0.02084 |
| Opus 5 | $0.00072 | $0.01042 |
| Sonnet 5 | $0.00029 | $0.00417 |
| Haiku 4.5 | $0.00014 | $0.00208 |
Grade A, and why
eval-harness 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 11d 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 — 217 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/eval-harness — Deterministic structural gate for skill outputs
Stdlib Python that scores a markdown artifact against a declarative rubric.json. Runs in <100ms per artifact. No LLM calls — pure regex / section extraction / count checks.
The gate is complementary to LLM reviewers, not replacement. Structural rules are deterministic and cheap; voice / brand / taste judgment stays in voice-reviewer, design-reviewer, skill-reviewer.
When to use
- Pre-commit hook on
.claude/skills/PRs — catch missing sections, malformed frontmatter, banned keywords before the LLM reviewer pass - CI gate on client deliverable PRs (proposals, positioning docs, messaging libraries, landing-page copy, AEO articles)
- Local lint during authoring:
python run.py --skill positioning - Sanity-check before voice-reviewer runs (saves expensive LLM passes on structurally broken drafts)
When NOT to use
- For qualitative judgment (tone, voice, brand fit) — that's
voice-reviewer - For visual / design output — that's
design-reviewer - For skill-internal SKILL.md quality — that's
skill-reviewer - For one-off outputs without a stable rubric — rubric authoring costs more than the gate saves below ~5 artifacts
Anatomy
eval-harness/
├── SKILL.md ← this file
├── engine.py ← check evaluator (stdlib, no LLM)
├── run.py ← walker + report renderer
├── NOTICE.md ← attribution to source pattern
├── the premium reference
│ ├── check-types.md ← every check type with example
│ └── rubric-authoring.md ← how to write a rubric for a Genesys skill
└── rubrics/
├── client-proposals.json
├── positioning.json
├── messaging.json
├── landing-page-copy.json
└── aeo-content.json
Workflow at a glance
| Step | Action | Output |
|---|---|---|
| 1 | Author rubrics/{skill}.json per gated skill |
Declarative criteria with weights + severity |
| 2 | Run python run.py --skill {skill} against latest artifact |
Per-criterion pass/fail + weighted score |
| 3 | On PASS, advance to LLM reviewer pass | voice-reviewer / design-reviewer / skill-reviewer |
| 4 | On FAIL with blocker, fix structure first | Re-run step 2 |
| 5 | On FAIL with only warnings, surface but proceed | Warnings logged, not blocking |
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.
- 11d ago First seen · 217 lines · 144 tokens per session scan A 246a81835bb3
eval-harness is a skill published in the GitHub repository matteotitta/genesys-skills (36 stars, last pushed 1mo ago), licensed MIT. It adds 144 tokens to every session and 2,084 once invoked, about $0.0007 per session on Opus 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
gingiris-b2b-growth
🇺🇸 B2B SaaS Growth — PLG vs SLG Playbook — Diagnose whether your problem is distribution, pricing, or PMF. PLG/SLG selection by ACV and sales cycle, the 5-stage path from $0 to $10M ARR, NRR discipline, affiliate & channel motion, enterprise tiering. Built from HeyGen, Deel, Vercel, Supabase, Snowflake patterns.…
gr-b2b-growth
A guide to growing a business-to-business software product from early user research to large-scale sales. B2B software is sold to companies rather than individual consumers.
go-to-market-playbook
A reusable Go-to-Market strategy template for both B2B and B2C launches. Covers positioning, messaging, ICP definition, channel selection, and competitive analysis frameworks. By @WeiYipei.
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
playwright-cli
Automates browser interactions for testing and validating your own web applications using playwright-cli. Use when you need terminal-first browser control for navigation, form filling, screenshots, tracing, bound browser sessions, debugging, or generating Playwright test code. Only use against applications you own or…