Cue for AI

The cue-pitch skill.

One downloadable skill that teaches your AI to pitch like a publicist through Cue — research contacts, build brief-matched catalogues, fire carefully, work the inbox, and close licensing deals.

What it is

cue-pitch is an Agent Skill — a single SKILL.md you load into your own AI (claude.ai, Claude Code, or any agent that reads skills). It carries the whole outreach playbook: how to research the right writers and supervisors, how to build per-recipient angles, and how to drive Cue's MCP tools to assemble, send, and track — in your workspace's voice, read live at runtime. One skill, every workspace; nothing to fork or configure.

The division of labor is the whole idea: research is your AI's; careful firing is Cue's. Your AI finds the right ten people and writes the angle. Cue enforces verification, suppression, caps, warmup, and send windows in code — for every send, no matter what any AI asks for.

Get it

Download cue-pitch.zip or read the raw SKILL.md →

claude.ai — Settings → Capabilities → Skills → upload cue-pitch.zip.
Claude Code — unzip into ~/.claude/skills/ so you have ~/.claude/skills/cue-pitch/SKILL.md.
Anything else — it's plain markdown with frontmatter; paste it wherever your agent takes standing instructions.

Then connect the Cue MCP server (see Cue speaks MCP) and ask your AI to pitch something. The outreach tools ride the outreach campaigns addon on the Publicist plan.

Better: download it from inside Cue. The "your AI" button on your pitches, EPKs, outreach, and contacts pages hands you this same skill personalized to your workspace — your release ids, pitch ids, what each EPK is still missing — so your AI skips the warm-up questions. The download above is the generic edition.

What your AI learns

Ground itself in your press kit and catalog before writing a word — your voice, not its own.
Research contacts to a quality bar (ten right people beat a hundred maybes) and import them with consent honesty.
Press track: pitch pages with per-writer, proof-of-research openers — a generic "love your work" is banned by rule.
Sync track: briefs → per-track metadata enrichment → explainable catalogue assembly → a /c page the supervisor can play and filter.
Careful firing: preflight, your explicit yes, then launch — inside Cue's caps and business-hours windows.
Inbox triage by intent, same-day reply-to-reassemble on new briefs, license leads, and pre-filled cue sheets.

The rules it carries

Some of these Cue enforces in code for every send — suppressions, the three-touch cap, stop-on-reply, honest one-stop rendering. The rest the skill imposes on your AI as standing rules, so it cooperates with the gate instead of fighting it:

Money, rights, and clearance answers are always yours — the AI drafts, you commit.
One-stop is only shown when you've confirmed the rights; unconfirmed reads "contact to confirm".
No fabricated quotes, press, or numbers; suppressions and unsubscribes are permanent.
Three touches max, absolute stop-on-reply, and nothing launches without your explicit yes.