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 joshuawheelock/grill-me --skill grill-megit clone --depth 1 https://github.com/joshuawheelock/grill-meWrote 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/joshuawheelock/grill-me/grill-me)<a href="https://agentmods.dev/skills/joshuawheelock/grill-me/grill-me"><img src="https://agentmods.dev/badge/skills/joshuawheelock/grill-me/grill-me/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/joshuawheelock/grill-me/grill-me"><img src="https://agentmods.dev/badge/skills/joshuawheelock/grill-me/grill-me.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 | $0.00093 | $0.00795 |
| Opus 5 | $0.00046 | $0.00398 |
| Sonnet 5 | $0.00019 | $0.00159 |
| Haiku 4.5 | $0.00009 | $0.00080 |
Grade A, and why
grill-me 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 12d 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 — 33 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Grill Me
Test the developer's working mental model of the codebase. Prefer questions about behavior, control flow, data flow, state, boundaries, tradeoffs, and failure handling over syntax or trivia.
Run the quiz
- Treat the current worktree as authoritative and keep the quiz read-only.
- Locate this skill's directory and select its native
bin/pick-quiz-file-<os>-<arch>executable, adding.exeon Windows. Usewindows,darwin, orlinuxfor the OS and mapx86_64toamd64andaarch64toarm64. Run it with the current repository root. It prints one random repository-relative path and requires no language runtime. For later questions, pass every previously selected path as a separate--exclude-pathargument. Do not enumerate the repository in model-visible output before running the helper. - Read the selected file or a relevant bounded section. Choose a substantive behavior or symbol, then use exact-symbol searches to inspect only the callers, callees, tests, configuration, and downstream effects needed to establish the answer. Treat the selected file as a starting point, not the complete source of truth.
- Rerun the helper when a selected file cannot support a useful question. After three unproductive selections—or when the current platform is unsupported, executable permission is denied, or no viable candidates are available—fall back to inspecting enough of the repository to build a private pool of viable subjects from distinct components. Exclude generated code, vendored dependencies, and facts that amount only to recalling a name or line number.
- Track selected paths and question subjects in the current conversation. Sample without replacement until the available subjects are exhausted.
- Form one focused, self-contained question about how that part of the code works. Ask for the mechanism, consequence, or rationale encoded in the implementation. Keep the answer and supporting evidence private. Do not ask multiple independent questions in one turn.
- Wait for the user's response. Do not provide hints unless the user explicitly asks for one; a hint does not resolve the question.
- Grade the response semantically against the inspected code. Accept different terminology when the explanation preserves the actual behavior and important consequences. Do not require details the question did not ask for.
- Resolve the question using exactly one of these outcomes:
- If correct, reply
Correct. Would you like another question? - If incorrect, begin with
Incorrect.Then concisely explain the actual behavior and cite the relevant local files or symbols. End withWould you like another question?
- If correct, reply
- If the user continues, repeat from file selection and ask exactly one new question. If the user challenges the grade, re-open the relevant implementation, address the challenge, and correct the grade when warranted before offering to continue.
What ships with it
8 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.
- 12d ago First seen · 33 lines · 93 tokens per session scan A d2f6d29c366c
grill-me is a skill published in the GitHub repository joshuawheelock/grill-me (6 stars, last pushed 21d ago), licensed MIT. It adds 93 tokens to every session and 795 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
spar
Make the user predict before you implement, then show them where their model was wrong and log the gap. Use whenever you are about to write or edit source code in a project the user is learning, such as adding an endpoint, service, data model, migration, background job, or wiring up a dependency. Also use when the…
pop-quiz
Use when the user wants to be quizzed or tested on work that was just built for them. Triggers include "quiz me", "test my understanding", "do I actually get this", "pop quiz on what you built", "test me on this branch". Not for explaining or re-teaching the work — only when they want to be tested on it.
reading-metaskill
A reading and learning guide based on building a regular reading habit, choosing books, and understanding difficult subjects through original works and explanation.
textbook-distillation
Turn a textbook or long-form source into a self-paced learning track: chapter map, lesson plan, then self-contained HTML lecture notes (styled as the human specifies) with worked examples, exercises, and checkpoint questions. Read this for "lecture notes" / "study notes" / "course" / "syllabus" requests from a…
blind-spot-pass
Use before starting work in a domain you don't know well, to surface the "unknown unknowns" — the things you don't even know to ask about — and learn just enough to prompt and decide well. Implements the "blind spot pass" pattern from Anthropic's Fable "finding your unknowns" field guide. Triggers when you say "I'm…
prompts-chat
Use when searching, installing, or improving AI skills and prompts via prompts.chat or skills.sh. Triggers on skill search, prompt lookup, install skill, improve prompt, prompts.chat.