suede-ship-copy

Orchestrates a multi-agent DAG to research, fact-audit, draft, and gate publish-ready marketing copy.

123|10|Updated May 24, 2026
One-click install
npx skills add https://github.com/JasonColapietro/suede-creator-skills --skill suede-ship-copy-jasoncolapietro
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: suede-ship-copy
Source: https://github.com/JasonColapietro/suede-creator-skills/tree/main/skills/suede-ship-copy
Command: npx skills add https://github.com/JasonColapietro/suede-creator-skills --skill suede-ship-copy-jasoncolapietro

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? High-stakes public copy (landing pages, launch posts, emails, ads) often contains invented facts, unsupported claims, and AI slop that erode trust. This Skill runs a structured multi-agent workflow that researches sources, audits every agent-generated claim against evidence, and produces a fact-checked draft with an evidence record. ## Core Features & Use Cases - Claim audit pipeline: Five blind research lenses gather provenanced facts, then an audit phase verifies each agent-generated claim as holds, overstated, unsourceable, or stale before any writer may assert it. - Message-ownership drafting: A planner assigns one message per section, disjoint writers draft in parallel, and an assembler unifies voice while preserving protected strings and author placeholders byte-exact. - Adversarial review and readiness gate: Four review lenses, dual refutation verifiers, a Slop Stop cleanup pass, deterministic checks, and a read-only publish-readiness verdict with an evidence record. - Use Case: Ask it to write a launch landing page for your product, supply the repo and pricing config as sources plus your own stated facts as given, and receive a publish-ready draft with placeholders for anything no source supports. ## Quick Start Use suede-ship-copy to write a landing page for my product launch, using the repo and pricing page as sources, and ask me which agent budget to run before starting.

Frequently Asked Questions about suede-ship-copy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write fact-checked landing page copy with AI agents?

Invoke the workflow with a piece description, target surface, and source paths or URLs. The DAG runs research lenses, audits every agent-generated claim against its source, drafts sections in parallel, and returns a draft plus an evidence record with placeholders for unsourced facts.

What is the difference between suede-ship-copy and suede-copy?

suede-ship-copy runs a research-heavy multi-agent DAG with claim auditing and adversarial review for high-stakes pieces strangers will read. suede-copy handles one surface in a single pass when facts are already established and no research is needed.

Does suede-ship-copy publish or edit live pages?

No. It writes exactly two files to a draft output directory: the draft and the evidence record. An output path pointing at already published copy halts the run, and it never posts, sends, commits, or deploys anything.

How many agents does a suede-ship-copy run use?

Runs range from about 30 agents at the light budget to 50 at deep, with standard at 38-45. The user must choose the budget before launch; sections are never cut to fit, only research and verification depth scales.

Can I supply my own facts that should not be audited?

Yes. Pass facts as the `given` argument or state them in the brief. They enter the permitted claim set marked as user-origin, skip the audit entirely, and no phase may verify, hedge, or gate on them.

Why did my suede-ship-copy run halt with a collision error?

A section map collision means two sections own the same message, a section cites a claim that failed the audit, a protected string is unassigned or double-assigned, or section budgets exceed the word budget. Fix it by re-planning: merge sections, supply sources, or adjust the budget.