Proseify - Anti-Prose-Slop Setup Guide¶
Source: proseify.xyz via skills.sh - 37,577 installs (anti-prose-slop); first seen Oct 8, 2026 (trending pass; the source is the product site itself, sized via the skills.sh API)
Skill file: proseify.xyz/skills/anti-prose-slop/SKILL.md (MIT, fetchable for review - verified Oct 8, 2026)
MCP server: Dionisselami/proseify-mcp (MIT; published in the official MCP registry as io.github.Dionisselami/proseify-mcp)
Category: Writing / Creative Content
Quality Tier: 🔵 Community (single-vendor product skill distributed from the product site; MIT; no skills.sh security verdicts published - due diligence required)
Proseify is a hosted MCP book-writing engine: give it a one-line premise and it picks a genre recipe, pulls model passages from a curated public-domain corpus, plans the outline, drafts chapter by chapter, and scores its own work. The anti-prose-slop skill is the workflow layer - discipline for using those tools well (genre recipe first, corpus grounding, chapter structure, and a per-chapter anti-slop editing pass) instead of raw tool access alone.
The skill's own compatibility line names Hermes among the agents it works with: "Works with any AI agent that can call the Proseify MCP server (Claude Code, Codex, Cursor, OpenCode, Hermes)."
Installation¶
1. Connect the MCP server¶
Sign in at proseify.xyz and get a key (plans start at $9/month; a one-time lifetime option also exists). Then add the server to your agent's MCP config:
{
"mcpServers": {
"proseify": {
"type": "http",
"url": "https://mcp.proseify.xyz/mcp",
"headers": { "Authorization": "Bearer YOUR_KEY" }
}
}
}
2. Install the skill¶
npx skills add https://proseify.xyz --skill anti-prose-slop -y
The skill has no value without the MCP connection: no key, no tools.
What It Provides¶
| Tool | What it does |
|---|---|
list_genres |
11 genres with a growing library of books (theatre, horror, romance, adventure, literary, and more) |
search_corpus |
Full-text search across all chapters with FTS5 syntax (quotes, AND/OR/NOT, wildcards) |
get_genre_recipe |
Pacing beats, dialogue ratio, and stylistic anchors derived from the genre's tradition |
get_style_references |
Model passages from books matching a genre - the register you are aiming for |
plan_book |
Chapter-by-chapter outline from a one-line premise |
write_book |
The one-shot flow: plan, draft, evaluate in one session |
evaluate_book |
Scores a drafted book against genre benchmarks so weak chapters get revised |
The corpus is drawn from Project Gutenberg and similar public-domain collections, verified against original headers (the repository describes 81 classics across 11 genres); the project states commercial use of output built on it is safe.
The skill's anti-slop checklist targets the real failure mode of AI prose - sameness: "seemed to", "began to", stacked adverbs, repeated sentence constructions, filter words, dialogue-as-exposition, and chapter openings that always start with weather.
Why This Matters for Hermes Agents¶
Longform creative work is where generic agents fail most visibly: endless prose in a single register. This pipeline forces a genre-first order of operations (recipe, corpus grounding, outline, drafts, scored revision) and gives the agent a scoring loop over its own output. It is also a clean example of the skill-on-top-of-MCP pattern: the MCP server supplies tools, the skill supplies workflow discipline, and the combination produces a manuscript a human can edit and ship.
Usage¶
| You say | What happens |
|---|---|
| "I want a gothic horror novel about a lighthouse keeper who finds a body that should not exist. 32 chapters." | Genre recipe, corpus passages, outline, then chapter drafts with scored revision |
| "Find me the register for a Victorian adventure story" | search_corpus plus get_style_references return model passages from the tradition |
| "Score this draft against genre norms" | evaluate_book returns a benchmark comparison the agent uses to revise |
| "Plan the book before writing anything" | plan_book produces the full chapter outline for review first |
Verification¶
# Confirm the skill is installed
npx skills list | grep anti-prose-slop
# Confirm the MCP endpoint responds (expect an auth challenge without a key)
curl -s -o /dev/null -w "%{http_code}" https://mcp.proseify.xyz/mcp
Security¶
No skills.sh content-security verdicts were published for this skill at verification time (the source is the product site, and its skills.sh detail page is not available). What is verifiable:
| Item | Status |
|---|---|
| Skill license | MIT (stated in the skill frontmatter) |
| Skill file | Fetchable and reviewable before install: https://proseify.xyz/skills/anti-prose-slop/SKILL.md |
| MCP server repo | Dionisselami/proseify-mcp, MIT |
| External service | Drafting runs on Proseify's hosted service; your premise and drafts go to their API |
Review the skill file and the service's terms before connecting; the skill instructs the agent to write books via the hosted API with your key.
Related¶
- Book-to-Skill - Book to Agent Skill Converter Setup - turns books into agent skills; the inverse direction of this pipeline
- Writing Plans - Subagent Development Setup - planning discipline for long agent tasks
- Skills Catalog | Skills Marketplace
Pro Tips¶
- Connect the MCP first. The skill explicitly depends on it - installation order matters.
- Budget before you generate. Set a chapter count and word budget in the ask; the pipeline plans against it.
- Read the genre recipe yourself once. It explains why the output reads the way it does; it is short and makes revisions easier.
- Do not fight the score loop. Let
evaluate_bookgrade first drafts and revise the weak chapters; that iteration is the product. - Keep the corpus claim in scope. Public-domain grounding covers the reference library; your manuscript and its commercial use are your own responsibility under the service terms.
← Skills Catalog | Skills Marketplace → Powered by CorpusIQ
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.