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 skills/recusive/skills/problem-clarifiernpx skills add Recusive/Skills --skill problem-clarifiergit clone --depth 1 https://github.com/Recusive/SkillsWhat 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 | $0.00108 | $0.01590 |
| Opus 5 | $0.00054 | $0.00795 |
| Sonnet 5 | $0.00022 | $0.00318 |
| Haiku 4.5 | $0.00011 | $0.00159 |
Grade A, and why
problem-clarifier 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 2d 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Problem Clarifier
You are about to help someone articulate a problem they're experiencing. Your job is NOT to solve it, NOT to investigate code, and NOT to start planning. Your only job right now is to understand exactly what the problem is by reflecting it back until you and the user are on the same page.
Why this matters
When someone describes a problem in natural language, the words they use rarely capture the full picture. They might say "it's not turning blue" when the real issue is "it turns blue but then reverts on the next keystroke." If you plan based on the first description, you'll build a plan for the wrong problem. That wastes everyone's time and erodes trust.
The fix is simple: before planning anything, play back what you heard and let the user correct you. Do this until they say "yes, that's exactly it."
The Reflect-Back Loop
This is the core mechanic. It works like this:
1. Listen to the initial description
The user will describe their problem. Read it carefully. Pay attention to:
- What they're trying to do (the intent)
- What they expected to happen
- What actually happened instead
- Any specific triggers or steps they mention ("when I press space", "after I click submit")
2. Explore the relevant code
Before reflecting back, quickly explore the codebase to understand the area the user is talking about. This makes your reflect-back much more precise — you can use actual component names, function names, and file paths instead of vague descriptions. The user doesn't need to know the exact file path — that's your job.
Use Glob, Grep, and Read to understand the relevant code area. Keep this fast — you're building context, not debugging.
3. Reflect back with precision
Now restate the problem in your own words, using specific technical details from the code. Use the AskUserQuestion tool with a detailed restatement and ask if you've got it right.
The restatement should be:
- Specific: Include the actual UI element, component, file, or flow involved
- Step-by-step: Describe the exact sequence of user actions and what happens at each step
- Clear about the gap: State what SHOULD happen vs what DOES happen
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.
- 2d ago First seen · 120 lines · 108 tokens per session scan A 29ec0dfe986a
problem-clarifier is a skill published in the GitHub repository Recusive/Skills (2 stars, last pushed 4mo ago), licensed MIT. It adds 108 tokens to every session and 1,590 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
chinese-documentation
中文文档排版参考——中英文空格、全半角标点、术语保留、链接格式、中文文案排版指北约定。仅在用户显式 /chinese-documentation 时调用，不要根据上下文自动触发。.
chinese-git-workflow
国内 Git 平台配置参考——Gitee、Coding.net、极狐 GitLab、CNB 的 SSH/HTTPS/凭据/CI 接入差异与镜像同步配置。仅在用户显式 /chinese-git-workflow 时调用，不要根据上下文自动触发。.
chinese-code-review
中文 review 沟通参考——话术模板、分级标注（必须修复/建议修改/仅供参考）、国内团队常见反模式应对。仅在用户显式 /chinese-code-review 时调用，不要根据上下文自动触发。.
dispatching-parallel-agents
当面对 2 个以上可以独立进行、无共享状态或顺序依赖的任务时使用.
executing-plans
当你有一份书面实现计划需要在单独的会话中执行，并设有审查检查点时使用.
verification-before-completion
在宣称工作完成、已修复或测试通过之前使用，在提交或创建 PR 之前——必须运行验证命令并确认输出后才能声称成功；始终用证据支撑断言.