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 agentmods add commands/florini/florians-claude-code-kit/grill-megit clone --depth 1 https://github.com/FlorinI/Florians-Claude-Code-KitWrote 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/commands/florini/florians-claude-code-kit/grill-me)<a href="https://agentmods.dev/commands/florini/florians-claude-code-kit/grill-me"><img src="https://agentmods.dev/badge/commands/florini/florians-claude-code-kit/grill-me.svg" alt="Measured on agentmods" 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 | $0.00042 | $0.00813 |
| Opus 5 | $0.00021 | $0.00407 |
| Sonnet 5 | $0.00008 | $0.00163 |
| Haiku 4.5 | $0.00004 | $0.00081 |
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 4d 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 — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/grill-me
Pin down a feature through a one-question-at-a-time interview until we share an understanding of it. The output of this skill is alignment, not code — you do not write the implementation plan until I explicitly say "alignment complete".
The contract
- One question per turn. Never batch. Ask, wait for my answer, then ask the next. A wall of questions defeats the point.
- Recommend, don't just ask. Every question carries your recommended answer and the trade-off behind it — the cost of going the other way. I'm reacting to a proposal, not filling in a blank form. Make the recommendation concrete enough that "yes" is a complete answer.
- Adapt. Let my answer reshape what you ask next. If an answer closes a branch, skip its follow-ups; if it opens a risk, dig there. Don't read from a fixed script.
- Track the gaps. Keep a running sense of which areas below are still open, and steer toward the load-bearing unknowns first — the answers that, if they went the other way, would change the whole shape.
- Hold the line on the plan. No implementation plan, no code, no file layout, no step list until I say the magic words. If I ask for the plan early, point back to the open areas and ask whether I want to align them first or cut the interview short.
Areas to cover
Work through these, but in the order the feature demands — not top-to-bottom by rote:
- User goal — who is this for and what are they actually trying to accomplish? What does success look like from their side?
- Scope — the smallest version that delivers the goal. What's explicitly in for v1.
- Edge cases — empty/error/concurrent/large-input states; what happens when the happy path doesn't hold.
- Data model — what's stored, its shape, where it lives, migrations/back-compat.
- UI — surfaces, states, affordances; what the user sees and touches (skip fast if headless).
- Tests — what proves it works; the cases worth pinning down before building.
- Rollout — how it ships: flag, migration, phased, big-bang; how it's reverted if wrong.
- Risks — what could break, regress, or surprise us; the assumptions that would sink the design if false.
- Out-of-scope — what we are deliberately not doing, stated out loud so it doesn't creep back in.
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.
- 4d ago First seen · 54 lines · 42 tokens per session scan A 9c021e3552d0
grill-me is a command published in the GitHub repository FlorinI/Florians-Claude-Code-Kit (5 stars, last pushed 6d ago), licensed MIT. It adds 42 tokens to every session and 813 once invoked, about $0.0002 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 commands, from other repositories
sounds
List available notification sounds and preview them.
notification
Use the unified toast notification system for all user-facing notifications in Conduit.
dev
You are continuing development on Conduit, an AI-powered remote connection manager. Follow this workflow precisely.
run
显式启动一次隔离的 Windows Computer Use 任务.
notifications-settings
Alias for settings — Configure notification sounds and settings.
cross-platform
Command "cross-platform" from Awesome-Embedded-Learning-Studio/CFDesktop, covering /cross-platform — 跨平台兼容性检查, 触发方式, 支持平台, 检查流程 and step 1: 条件编译检查.