outline

name	outline
description	Make an outline for a post.

Outline

Use this skill as the shell around the outline-generator subagent.

This skill does not own outline design principles. Keep outline methodology in the outline-generator subagent. The skill only prepares context, invokes the subagent, and interprets the response for the user.

Responsibilities

• Summarize the current conversation and any source notes.
• Form an append-only spawning prompt for outline-generator.
• Invoke the outline-generator subagent with the subagent tool and that prompt.
• Read the subagent response.
• Present the useful output to the user and ask what to change next.

Limits

MUST NOT:

• You MUST NOT define outline structure rules in this skill.
• You MUST NOT duplicate outline methodology here.
• You MUST NOT draft the article body unless the user explicitly asks.
• You MUST NOT create or edit the post file unless the user explicitly asks.
• You MUST NOT treat the subagent as a state machine.

Context Summary

Before invoking outline-generator, create a compact summary from the current conversation.

Include only information the subagent needs:

1. Core thesis: What is the article trying to say in one sentence?
2. Personal stake: Why does this matter to the author specifically?
3. Reader promise: What will the reader understand better after reading this?
4. Key raw materials: Important experiences, arguments, examples, quotes, contrasts, or observations.
5. Tension / conflict: What is the central disagreement, paradox, or unresolved problem?
6. Constraints: Tone, register, audience, things to avoid, authenticity rules.
7. Existing post: The path to the existing post. Optional. Empty when the post is not created.
8. Revisions: The historical revisions of the discussed outlines.

Invoke

Call the subagent tool like this:

{
  "agent": "outline-generator",
  "task": "<spawning prompt>"
}

Spawning Prompt

Pass the summary to outline-generator as an append-only input bundle.

Use this prompt shape:

<SPAWNING_PROMPT_TEMPLATE>

## Core thesis
<!-- What is the article trying to say in one sentence? -->
## Personal Stake
<!-- Why does this matter to the author specifically? -->
## Reader Promise
<!-- What will the reader understand better after reading this? -->
## Key Raw Materials
<!-- Important experiences, arguments, examples, quotes, contrasts, or observations. -->
## Tension / Conflict
<!-- What is the central disagreement, paradox, or unresolved problem? -->
## Constraints
<!-- Tone, register, audience, things to avoid, authenticity rules. -->
## Existing Post
<!-- The path to the existing post. Optional. Empty when the post is not created. -->
## Revisions
<!-- The historical revisions of the discussed outlines. You **MUST** list in the following format:

### 1. [Design purpose of the outline V1]

**Section Titles:**

1. <Section Title 1. Verbatim copy from the `outline-generator` subagent outputs.>
2. <Section Title 2. Verbatim copy from the `outline-generator` subagent outputs.>

**Writing Approach:**

1. <Section Title 1>: <Writing approach of       ``  >
   
   
 : 
   
   



 ... 
 ... 

---





 
 



 : 
   
   
 : 
   
   



 ... 
 ... 

-->

</SPAWNING_PROMPT_TEMPLATE>

MUST:

You MUST explicitly state "The user has no opinions for this version." in User's Opinions section if the user found no opinions for the particular version of outline. You MUST exhaust the list of the user's opinions for each version of the outline.

MUST NOT:

You MUST NOT write prose. You MUST NOT invent examples. You MUST NOT smooth out the author's voice.

Response Handling

MUST:

You MUST honestly present what the outline-generator subagent responded. You MUST unwrap the outer markdown codeblock of the response before outputing it. You MUST include all the sections in ONE SINGLE markdown codeblock.

MUST NOT:

You MUST NOT include each section in SEPARATE markdown codeblocks.