Skip to content
Agent Orchestration
Skill

/discussion

Have an interactive discussion about a topic, approach, feature, or bug without making code changes. Researches the codebase as needed and talks through options. Use to think through an approach before planning or coding.

BOOST
From plugin
pane
54537 skills4 agents
Install
$ npx -y skills add greenfield-inc/Pane --skill discussion --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/discussion

Context preview

The summary Claude sees to decide when to auto-load this skill.

Have an interactive discussion about a topic, approach, feature, or bug without making code changes. Researches the codebase as needed and talks through options. Use to think through an approach before planning or coding.

SKILL.md

discussion.SKILL.md
name: discussion
description: Have an interactive discussion about a topic, approach, feature, or bug without making code changes. Researches the codebase as needed and talks through options. Use to think through an approach before planning or coding.
argument-hint: "[topic or question to discuss]"

Discussion

Topic: $ARGUMENTS

Talk the topic through with the user, back and forth, until the ideas, tradeoffs, and goal are clear enough to plan or build.

Conversation only

Read code and research freely. Your only output is conversation. Never edit, create, or delete project files, and never propose diffs or patches to apply.

1. Research as needed

  • Clarify the topic and the outcome the user wants.
  • When the topic depends on current behavior, read the relevant code paths.

Use the `explorer` subagent or `codebase-explorer` for code, and `research-web` for external libraries or approaches, including mid-conversation when new questions come up.

  • Research only what the conversation needs.

2. Discuss with the user

  • Apply the `rewrite-simply` skill to every discussion response before you send

it.

  • Present findings, first thoughts, concrete options, constraints, and

tradeoffs.

  • Ask targeted questions about preferences, constraints, and goals.
  • Be opinionated. Recommend with reasons, separate fact from recommendation,

and defer to the user's judgment.

Prototype before you ask

Classify every fork question before you raise it. If running something can answer it (behavior, timing, output, performance, layout), build the cheapest throwaway probe outside the project tree, run it, and present the result with a recommendation. Save questions for product or preference calls that no experiment can settle. A probe usually answers faster than a person, and it gives them a result to react to.

3. Hand off and suggest next steps

Recap in the conversation:

  • the trigger and why it matters
  • the desired outcome
  • constraints and non-goals
  • decisions and open questions
  • available source references

Mark which points are the user's intent and which are your assumptions. Then recommend the next step, usually `create-ticket` when the work is ready to delegate, `options` when more than one approach is live, `brief` for a write-up, `investigate` for a bug, or implementation.

Suggested next steps:
- `/create-ticket [description]` - Capture the work for delegation
- `/options [question]` - Lay out the approaches and trade-offs
- `/discussion [follow-up]` - Continue exploring a specific aspect
- `/research-web [topic]` - Look up external documentation

Topic to discuss: $ARGUMENTS

Read more
Ships withpane

pnpm dlx runpane@latest

Get the whole plugin

Other skills on pane.