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 yuyou-dev/GrillMeJewel --skill grill-me-jewelgit clone --depth 1 https://github.com/yuyou-dev/GrillMeJewelWrote 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/yuyou-dev/grillmejewel/grill-me-jewel)<a href="https://agentmods.dev/skills/yuyou-dev/grillmejewel/grill-me-jewel"><img src="https://agentmods.dev/badge/skills/yuyou-dev/grillmejewel/grill-me-jewel/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/yuyou-dev/grillmejewel/grill-me-jewel"><img src="https://agentmods.dev/badge/skills/yuyou-dev/grillmejewel/grill-me-jewel.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00108 | $0.01341 |
| Opus 5.5 | $0.00043 | $0.00536 |
| Sonnet 5.5 | $0.00022 | $0.00268 |
| Haiku 4.5 | $0.00011 | $0.00134 |
Grade A, and why
grill-me-jewel 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 1mo 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 — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Grill Me Jewel
Purpose
Help a beginner discover and articulate what jewelry they want to design. Use adaptive discovery by default, or four purposeful discovery rounds for an explicit deep interview. Both modes end in a separate confirmation round. Preserve every answer, then use Codex image generation to generate the requested real design images.
Interview
- Read
references/design-frontier.md. Reuse facts from the conversation and attachments, then usemode: adaptiveto ask only unresolved decisions that change the result. Skip known facts. For an explicit deep interview, usemode: fulland complete foundation, meaning, design language, and variation/delivery as four submitted Apps UI rounds. - For every round, call
ask_grill_me_questionsas the primary interaction surface with the exactstageand sequentialround. Ask one to four currently answerable fields. Discover that exact tool name before falling back to prose. Use concise chat questions only after real tool discovery fails or the call errors. - Use stable lowercase field and option ids. Offer an
otheroption when a useful answer may fall outside the list. In foundation, askdelivery_countunless the user already supplied a count; offercount_1,count_2,count_4,count_8, and a custom value. Never ask for providers, concurrency, internal job ids, API keys, or cost. Distinguish image count from piece/pair and explicit stone/pearl counts inside each image. Flexible materials or style never increase the accepted delivery count. - After submission, summarize only the newly established facts and preserve all earlier answers. Continue with the next unresolved decisions in adaptive mode, or the next stage in full mode; never answer the user's side of the interview.
- In variation/delivery, split locked facts from flexible axes. For multiple outputs, define named candidate branches that each change at least three visible design axes while preserving product identity, wearing logic, and the central story.
- After the necessary decisions are resolved (four discovery rounds in full mode), present the assembled brief through one final
ask_grill_me_questionsconfirmation round withstage: confirmationand the next sequentialround(5 or greater in full mode). Ask whether to confirm it or revise it, with a text field for corrections when needed. Include the accepted counts, reference roles, visible translation of the story, candidate matrix, and Presentation Lock fromreferences/design-frontier.md. This uses the existing confirmation, not an extra round. Label creative defaults separately from user-supplied facts. - After confirmation, read
references/image2-generation.md, compile one production prompt per requested design, and invoke Codex$imagegen/ built-in image generation. Carry the complete accepted brief into every prompt; do not reduce it to a style summary or resume interviewing during generation. - Return the final brief in Markdown under: Objective, Product, Design Direction, Materials and Craft, Source Assets, Output Intent, Locked Facts, Flexible Details. Present every real generated image inline. Never claim an image exists unless the tool returned it.
What ships with it
5 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.
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.
- 1mo ago Changed · +21 lines · +3 tokens per session 5000e3f5fe08
- 2mo ago First seen · 69 lines · 105 tokens per session scan A d8593fa692bf
grill-me-jewel is a skill published in the GitHub repository yuyou-dev/GrillMeJewel (1 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 108 tokens to every session and 1,341 once invoked, about $0.0004 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 2026-08-31.
grill-me-jewel compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| grill-me-jewelyuyou-dev/GrillMeJewel | Skill | 1 | 1mo ago | Apache-2.0 | A | 108 | Codex |
| web-design-engineerConardLi/garden-skills | Skill | 15k | 13d ago | MIT | A | 96 | Codex |
| asset-setSateezg/codex-bridge | Skill | 244 | 3mo ago | MIT | A | 85 | Claude Code |
| claude-design-skill0xmhha/claude-design-skill | Skill | 3 | 3mo ago | Apache-2.0 | A | 184 | Claude Code |
| engineering-blueprint-sheetFANzR-arch/Phil-design-skills | Skill | 111 | 2mo ago | none | A | 266 | Claude Code, Codex |
| quiet-paper-specimenFANzR-arch/Phil-design-skills | Skill | 111 | 2mo ago | none | A | 198 | Claude Code, Codex |
| brand-icon-bannerFANzR-arch/Phil-design-skills | Skill | 111 | 2mo ago | none | A | 190 | Claude Code, Codex |