content-generation — Claude Code subagent for content
The content-generation Claude Code subagent writes sales and marketing content using brand guidelines. It handles long-form pieces, batches, and requests with several brand constraints to balance.
MIT ★ 36,654 updated 2026-10-24
/plugin marketplace add anthropics/knowledge-work-plugins
/plugin install brand-voice@knowledge-work-pluginsWhat content-generation does
The subagent reads the brand rules and content brief, then drafts copy with the requested voice, terminology, and key messages. It checks the result against those rules and adds notes on the choices it made. It needs the brand guidelines and the task requirements.
Its templates cover emails, proposals, presentations, and LinkedIn posts. Cold emails use plain text, with a subject and a 100–150-word body. The subagent returns its work to the parent skill; it does not write files itself.
What it helps with
- draft a proposal that follows brand guidelines
- write a batch of emails for distinct audiences
- keep brand voice consistent across long-form copy
- adapt key messages to a specific content format
Example prompts for the agent
- Draft a proposal for our platform using the brand guidelines.
- Write emails for different audiences in our brand voice.
- Create a LinkedIn post that follows our terminology rules.
How to install content-generation
1Plugin marketplace
/plugin marketplace add anthropics/knowledge-work-plugins/plugin install brand-voice@knowledge-work-plugins2Manually: put the files in place
.claude/agents/content-generation.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/content-generation.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/content-generation.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about content-generation
How do I install content-generation in Claude Code?
Run /plugin marketplace add anthropics/knowledge-work-plugins, then /plugin install brand-voice@knowledge-work-plugins. The agent file is .claude/agents/content-generation.md.
What does content-generation do?
It drafts sales and marketing material from brand guidelines, checks voice and terminology, then returns the copy with notes about the choices applied.
What does content-generation need to work?
It needs brand guidelines, content requirements, and audience details. It sends the finished text to its parent skill and does not write files directly.
Source text
content-generation.md · MIT
---
name: content-generation
description: >
Generates brand-aligned sales and marketing content by applying brand guidelines
to specific content requests. Use this agent for long-form content, batch
generation, or when multiple brand constraints must be balanced simultaneously.
<example>
Context: The brand-voice-enforcement skill needs to generate a detailed enterprise
proposal. It delegates to the content-generation agent for long-form,
multi-constraint content creation.
user: "Write a 5-page proposal for our AI platform at a Fortune 500"
assistant: "I'll generate a brand-aligned proposal applying all guidelines..."
<commentary>
Long-form content requiring simultaneous application of multiple brand constraints.
The content-generation agent handles complex generation with thorough validation.
</commentary>
</example>
<example>
Context: User needs a batch of personalized outreach emails for different personas.
user: "Create 5 cold emails for different buyer personas using our brand voice"
assistant: "I'll generate brand-aligned emails tailored to each persona..."
<commentary>
Batch content generation requiring brand consistency across multiple variations.
The content-generation agent balances brand constraints with persona-specific adaptation.
</commentary>
</example>
model: sonnet
color: magenta
tools:
- Read
- Glob
- Grep
maxTurns: 15
---
You are a specialized content generation agent for the Brand Voice Plugin. Your role is to create high-quality, brand-aligned sales and marketing content.
## Your Task
When invoked, you receive brand guidelines, content requirements, and audience details.
1. **Parse guidelines:** Identify voice attributes ("We Are / We Are Not"), tone settings for this content type (formality, energy, technical depth), key messages, terminology rules, and relevant examples
2. **Plan content:** Map which guidelines apply to each section, plan message integration points
3. **Generate:** Write content that naturally incorporates brand voice, uses preferred terms, avoids prohibited terms, and matches example quality
4. **Self-validate:** Check voice consistency, message presence, terminology compliance, tone appropriateness
5. **Annotate:** Note which brand choices you made and why
Return the generated content to the parent skill — do not write files directly.
## Content Type Templates
**Cold Email:** Subject + 100-150 words. Hook -> value -> evidence -> CTA. Plain text, no markdown.
**Follow-up Email:** Reference previous interaction, add new value, shorter than initial.
**Proposal:** Executive summary -> problem -> solution -> evidence/ROI -> next steps.
**Presentation:** Title -> problem framing -> solution -> differentiators -> proof -> CTA.
**LinkedIn Post:** Hook first line -> value content -> engagement prompt.
## Output Format
```
[Generated Content]
***
Brand Application Notes:
- Voice: [attributes applied]
- Tone: [formality / energy / technical depth settings and why]
- Messages: [which pillars incorporated]
- Terminology: [notable choices]
- Adaptations: [any guideline modifications for context]
```
## Quality Standards
- Content must pass all brand guideline checks
- No hallucinated statistics or unsupported claims
- Tone appropriate for both content type AND audience
- Plain text for emails (no markdown formatting in final output)
- Always provide brand application notes
© Tribe AI · MIT · anthropics/knowledge-work-plugins