Discuss
Agent skills for Claude.ai, Claude Code, ChatGPT, and Codex, covering document processing, study tools, and cross-agent workflows.
npx -y skills add maaarcooo/agent-skills --skill discussAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Ask clarifying questions and discuss direction before starting work. Use when the user invokes $discuss or says "discuss-first", wants to discuss requirements, clarify scope, or asks Codex to ask questions before proceeding with a task. Natural phrasings include "let's discuss before proceeding", "ask me questions for clarification or direction", and "let's discuss and make a plan first". Do not use for general requests to explain or discuss a topic when no pending task is being held back.
SKILL.md
2.4 KB, as published. Nobody here has run it
When invoked, treat this as the start of a genuine discussion about the task, not a clarification checklist.
Before doing any work:
- Ground yourself first, if the task touches existing code. When
the task references existing code, files, or project structure,
briefly inspect the relevant parts before asking anything, so your
questions cite real specifics ("you currently do X in
foo.ts, do you want this to change?") rather than abstractions. Keep the scan light — enough to ask informed questions, not a full audit. Skip this for greenfield or non-code tasks. - Share your initial read. State what you understand the goal to be, the assumptions you are making, and anything that looks like a key decision or trade-off. Engage with the specifics of this task.
- Ask focused questions that target real forks in direction. Prefer "do you want X or Y, because it changes Z" over generic intake questions. Skip anything already answered in the conversation or visible in the codebase.
- Attach a tentative opinion or recommendation where you have one, so I can react to it rather than answer cold.
Delivery style is your judgement. Default to one small batch of 3-5 questions. Go one or two at a time only when later questions genuinely depend on earlier answers.
Do not start the task until I have responded. After my answers:
- For complex or multi-step tasks, propose a concise plan and wait for confirmation.
- For simple tasks, proceed directly once clarified.
- If the user includes "plan" with the invocation (
$discuss plan), always propose a plan after gathering answers, regardless of task complexity. - Either way, when you proceed, state any remaining assumptions in one line (e.g. "proceeding with X and Y assumed") so nothing slips through silently.
Tone: conversational, like a collaborator thinking through the problem with me, not a form to fill.