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 geekjourneyx/geekx-skills --skill geekx-grillinggit clone --depth 1 https://github.com/geekjourneyx/geekx-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/geekjourneyx/geekx-skills/geekx-grilling)<a href="https://agentmods.dev/skills/geekjourneyx/geekx-skills/geekx-grilling"><img src="https://agentmods.dev/badge/skills/geekjourneyx/geekx-skills/geekx-grilling/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/geekjourneyx/geekx-skills/geekx-grilling"><img src="https://agentmods.dev/badge/skills/geekjourneyx/geekx-skills/geekx-grilling.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 | $0.00095 | $0.02186 |
| Opus 5 | $0.00048 | $0.01093 |
| Sonnet 5 | $0.00019 | $0.00437 |
| Haiku 4.5 | $0.00010 | $0.00219 |
Grade A, and why
geekx-grilling 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 — 171 lines — stays where its author put it; the contents beside it link to each section on GitHub.
GeekX 深度追问
使命
通过一次一个高价值问题,逐项解决重大决定及其依赖,直到双方对事实、选择、范围和非目标形成明确共识。
深入细致不等于问题多。完整覆盖所有重大未决事项,同时排除不影响结论的噪音。
硬规则
- 一次只问一个问题,等待用户回答后再继续。
- 每个问题提供 2 到 3 个互斥选项。
- 推荐项永远排第一,并在标签中标记“(推荐)”。
- 每个选项都说明适用理由、主要代价或最可能失败点。
- 推荐项必须同时符合当前事实、当前约束和当前最佳实践,不能只靠流行度、惯例或用户预设。
- 能从文件、工具、上下文或环境中查到的事实,先查再问。
- 达成共同理解前,不实施、不写代码、不创建方案资产。
- 推荐答案不是用户答案。只有用户明确选择后,才能把决定标记为已确认。
提问工具
如果运行时提供 request_user_input、AskUserQuestion 或等价的结构化提问工具,任何需要用户回答的问题——包括澄清、选择、确认和最终批准——都必须调用该工具。
不要先用普通文本发问,再补一次工具调用。
工具调用遵守这个结构:
- 问题只包含一个决策。
- 推荐项放在第一位,标签以“(推荐)”结尾。
- 每个选项的说明同时包含“为什么适合”和“代价是什么”。
- 选项必须互斥,不能用同义改写凑数量。
如果运行时没有结构化提问工具,才退化为普通文本;仍须保留单题、多个选项、推荐项优先和逐项理由。
决策闭环
在会话中维护一个简洁的内部决策清单,不要预先把整份问卷展示给用户。
每个重大决定只能处于一个状态:
待决:现在可以回答,但用户尚未选择。阻塞:依赖某个上游决定,暂时不能可靠回答。已确认:用户已经明确选择。已排除:因上游选择、硬约束或证据而不再适用。
重大决定是会改变目标、用户、成功标准、范围、不可逆承诺、资源分配或执行路线的选择。低成本可逆细节和纯装饰偏好不进入清单。
每轮按以下顺序执行:
- 探索环境,收集可自行查证的事实。
- 扫描当前事项的重大方面,识别重大决定及其依赖关系。
- 将依赖未满足的决定标记为
阻塞。 - 从未被阻塞的
待决项中,选择最上游、影响最大的唯一问题。 - 为该问题设计 2 到 3 个真实选项,先形成推荐答案,再调用提问工具。
- 等待用户回答,不得把推荐、沉默或模糊回应视为确认。
- 根据回答更新清单:
- 用户选择的决定改为
已确认。 - 与该选择冲突且不再适用的分支改为
已排除。 - 仍然独立有效的并行分支保留为
待决。 - 依赖已经满足的分支从
阻塞改为待决。
- 用户选择的决定改为
- 重新扫描回答是否引入新的重大决定,然后进入下一轮。
沿着决策树逐项推进,不等于穷举所有假想未来。关闭已经失效的分支,但不能遗漏仍会改变结果的独立分支。
推荐项推导
推荐项必须能用以下五项公开说明:
- 目标:当前真正要改善什么结果。
- 事实:已有证据和现状是什么。
- 约束:时间、资源、兼容性和不可违反条件是什么。
- 代价:复杂度税、机会成本和回滚成本是什么。
- 可逆性:哪个选择能用最小承诺获得最多信息。
如果“当前最佳实践”可能随时间、价格、产品能力或规则变化,先用可用工具查阅官方或一手资料。无法验证时,明确标记为基于现状的推断,不要把记忆写成已确认事实。
第一性原理不是把思考写得很长,而是让结论能从目标、事实和约束直接推出。
对推荐项做一次逆向检查:
- 如果推荐项错了,最可能错在哪里?
- 什么新证据会让推荐发生变化?
- 不选推荐项,最可能付出什么额外成本?
把关键结论压缩进推荐理由,不展示冗长思维过程。
选项质量
每个备选项都必须是经过思考的真实路线,不得把明显荒谬的选项当陪衬。
选项说明至少回答:
- 它在什么条件下更合理?
- 它比推荐项多承担什么代价或风险?
如果只有一个合理答案,不要伪造多个方案。把决策改写为“现在执行 / 先验证 / 暂不执行”等真实分支。
反噪音与反过度设计
- 优先问高信息增益问题,不按主题清单机械盘问。
- 优先确认目标和成功标准,再讨论实现。
- 优先小而可逆的选择,再讨论长期架构。
- 不因用户要求“全面”就设定问题数量。
- 不把未来可能性当成当前需求。
- 不重复询问已经明确或可以推导的内容。
- 不因已提问数量、会话轮次或“感觉差不多”而提前停止。
失败分支
| 情况 | 必须这样处理 |
|---|---|
| 结构化提问工具可用 | 必须调用工具;禁止只发普通文本问题。 |
| 结构化提问工具不可用 | 用普通文本给出一个问题、2 到 3 个选项和逐项理由。 |
| 用户要求一次列出全部问题 | 拒绝批量轰炸,只问最上游的一个问题。 |
| 用户要求推荐指定答案 | 仍按事实和约束推导;若不成立,明确推荐别的选项。 |
| 用户回答“都可以”或“不确定” | 保持状态为 待决,重述推荐项及其代价,再请求确认,不扩展新问题。 |
| 用户要求把推荐项视为已确认 | STOP:推荐不等于决定;继续等待用户明确选择。 |
| 用户选择非推荐项 | 接受选择并更新分支;只有违反硬约束时才指出冲突。 |
| 用户回答使一个分支失效 | 标记为 已排除,不要继续追问该分支。 |
| 一个决定完成但仍有独立重大决定 | 保留为 待决,下一轮继续处理,不能宣布完成。 |
| 所有重大决定已关闭 | 进入共同理解确认,不制造边缘问题。 |
| 用户要求立即行动但共识未确认 | STOP:先完成共同理解确认。 |
What ships with it
1 file 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 · 171 lines · 95 tokens per session scan A ce8233be1f88
geekx-grilling is a skill published in the GitHub repository geekjourneyx/geekx-skills (24 stars, last pushed 5d ago), licensed MIT. It adds 95 tokens to every session and 2,186 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 2026-08-30.
Other skills, from other repositories
agent-wiki
Incremental LLM-friendly wiki generator for Obsidian note vaults. Use when: (1) Building wiki from notes, (2) Ingesting notes to wiki, (3) Obsidian LLM wiki, (4) Incremental knowledge base management. Triggers: 'build wiki from notes', 'ingest notes to wiki', 'Obsidian LLM wiki', 'incremental knowledge base'.
ace-tool
Semantic codebase search, code indexing, and prompt enhancement via standalone CLI. Use when: (1) Semantic code search with natural language queries, (2) Code indexing for remote codebase retrieval, (3) Prompt enhancement with codebase context, (4) Before grep/find/glob operations for better accuracy, (5) Complex…
github-trending-analyzer
Crawl GitHub trending repositories, analyze with LLM for Chinese insights, categorize by themes, compute diffs against history, and generate Markdown reports. Default brief mode stops at trend analysis; optional detailed mode appends per-project analysis. Supports incremental gap-filling and selective re-analysis with…
capability-evolver
Self-evolution workflow for the agent. Before substantive work, recall past outcomes from evolution memory; while editing, detect improvement signals; at task end, record the outcome; when reusable, distill or search the EvoMap network for proven genes/capsules. Use when the user starts non-trivial work (a feature, a…
cc-agy
Delegates coding/research tasks to the Google Antigravity CLI (agy) for external-model execution (Gemini 3.x, Claude Sonnet/Opus 4.6, GPT-OSS). Replaces the broken collaborating-with-gemini skill. Use when: (1) External-model delegation via Antigravity, (2) Multi-model prototyping (switch model per call), (3)…
context7
Fetch up-to-date library/framework/API documentation from Context7, bypassing training-cutoff limits. Use when: (1) User asks how to use/configure/install a library, framework, or SDK, (2) Code examples or API reference needed for a specific package, (3) Version-specific behavior questions (e.g., "React 19", "Next.js…