// Personalize Sylph to your company. Ask the company name, then derive everything from the web - CONTEXT.md, brand voice, team, ICP, and standing instructions.
Generate a customer-specific enterprise service agreement by filling in the standard template with deal details.
Head of Data agent - creates reportings and ad-hoc analytics using nao MCP tools. Always use ask_nao to create conversations that persist in nao.
Review a contract against standard negotiation positions - flag deviations, generate redlines, provide business impact analysis. Use when reviewing vendor or customer agreements, when you need clause-by-clause analysis, or when preparing a negotiation strategy.
Create a new skill from scratch. Generates the SKILL.md, registers it in AGENTS.md, and sets up the content folder structure.
Set up an AI agent by validating her scope, schedule, and delivery channel. Run as `/sylph-setup-agent <agent-name>` or `/sylph-setup-agent all`.
Set up a specific Sylph skill by auto-pulling examples from the web, deriving insights, and configuring the skill. Run as `/sylph-setup-skill <skill-name>`.
| name | sylph-setup |
| description | Personalize Sylph to your company. Ask the company name, then derive everything from the web - CONTEXT.md, brand voice, team, ICP, and standing instructions. |
| Connector | Purpose |
|---|---|
| Web Search | Search for company info, team, socials |
| Web Fetch | Read company website, about page, blog, LinkedIn |
Ask the user as few things as possible. The only required input is the company name. Everything else should be derived from the web - website, LinkedIn, social profiles, blog, job postings, press coverage. Only ask the user to confirm and correct what you found.
Run /sylph-setup right after forking the repo, before using any other skill.
CONTEXT.md with derived company details.claude/MEMORY.md with initial preferencesbrand/guidelines/agents/Before anything else, welcome the user as the Chief Agent Officer (CAO) - the human who will review, approve, and have final say over everything the agents produce.
Ask:
Welcome, Chief Agent Officer. You're about to set up your company's AI brain. Before we start:
- What's your name?
- What are your pronouns? (she/her, he/him, they/them, etc.)
Store their name and pronouns in .claude/MEMORY.md so all agents and skills use them consistently.
Update AGENTS.md definition: replace the generic CAO description with their name and pronouns.
If the user's pronouns are she/her, add this message:
By the way - Sylph's creator Claire wanted me to tell you: she's happy to see a fellow tech sister on this repo. Welcome, Queen. 💜👸
Ask the user ONE question:
What's your company name? (and website URL if it's not obvious)
Then derive everything automatically:
From the web research, fill in every section of CONTEXT.md:
| Section | Source |
|---|---|
| What the company does | Website homepage + about page |
| The problem it solves | Website homepage + pricing page copy |
| How the product works | Docs, product page, or demo page |
| ICP | Pricing page, case studies, job descriptions |
| Team | About page, LinkedIn, careers page |
| Integrations | Docs, integrations page, product page |
| Company basics | Crunchbase, LinkedIn, footer of website |
| Key terminology | Docs, blog, product page |
| Standing instructions | Derive from brand voice observed on website + socials |
Show the user the draft CONTEXT.md and ask:
Here's what I found about [company]. Review this and tell me what to fix. I'll update it.
Only ask follow-up questions for things you genuinely couldn't find online (internal terminology, specific process rules, preferred tools).
Use the CAO's name in conversations (not "user" or "you"). Use their stated pronouns in all generated files.
[placeholder] markers remainAsk the user ONE question:
Any rules or preferences I should know about? (tool quirks, process rules, voice preferences, things to always/never do)
Write each answer as a dated entry in .claude/MEMORY.md. If she has nothing yet, that's fine - memory will grow as she uses skills.
Do not ask the user to describe her brand voice. Instead:
Generate brand/guidelines/voice.md from the analysis:
Show the draft to the user:
Here's the voice guide I derived from your existing content. What would you change?
Update based on her feedback.
Review agents/ and ask the user:
Which of these agents do you want active? (You can always activate more later.)
List the agents with one-line descriptions:
For each activated agent:
ROLE.md with company-specific boundaries (derived from CONTEXT.md)PROMPT.md with actual tool references (ask only for Slack channel IDs, calendar links, etc. that can't be found online)For deactivated agents:
Run a quick verification:
CONTEXT.md - does it fully describe the company?.claude/MEMORY.md - are initial preferences logged?ROLE.md and PROMPT.mdTell the user:
Your Sylph is initialized. Next step: run
/sylph-setup-skill linkedin(or whichever channel you want to start with) to set up your first content skill with real examples.