Masthead

Universal Editorial Standards

All four editors share these commitments:

1. Emdash elimination — Emdashes are a crutch. Replace them with proper punctuation: periods, semicolons, colons, commas, or parentheses. In rare cases, restructure the sentence entirely. The only acceptable emdash is in reported speech where it indicates an interruption.
2. Active voice — Prefer active constructions unless passive is genuinely better.
3. Precision — Vague claims must be made specific or cut.
4. No throat-clearing — Cut introductory filler. Start where the piece actually begins.
5. Earned length — Every paragraph must justify its existence. If it restates what came before, cut it.

Execution Pipeline

Phase 1: Read and Understand

Read the document. Identify:

• Genre (essay, report, memo, article, letter, proposal, etc.)
• Audience (specialist, general, internal, public)
• Approximate word count
• Current strengths and obvious problems

Store in hive-mind shared memory for all editors to reference.

Phase 2: Initialize the Hive-Mind

mcp__ruflo__hive-mind_init(queenId: "masthead-queen", topology: "mesh")

Spawn four specialist editors:

mcp__ruflo__hive-mind_spawn(count: 1, agentType: "specialist", prefix: "ed-ft", role: "specialist")
mcp__ruflo__hive-mind_spawn(count: 1, agentType: "specialist", prefix: "ed-economist", role: "specialist")
mcp__ruflo__hive-mind_spawn(count: 1, agentType: "specialist", prefix: "ed-lrb", role: "specialist")
mcp__ruflo__hive-mind_spawn(count: 1, agentType: "specialist", prefix: "ed-newyorker", role: "specialist")

Store the document content and analysis in shared memory:

mcp__ruflo__hive-mind_memory(action: "set", key: "document-text", value: "<full document>")
mcp__ruflo__hive-mind_memory(action: "set", key: "document-analysis", value: "<genre, audience, word count, initial observations>")

Phase 3: Independent Editorial Review

Each editor reviews the document from their publication's perspective. Run all four reviews in parallel using the Agent tool. Each agent should:

1. Read the document from shared memory
2. Produce a structured editorial report following their persona (see references/editor-personas.md)
3. Store their review in shared memory

mcp__ruflo__hive-mind_memory(action: "set", key: "review-ft", value: "<FT editor's review>")
mcp__ruflo__hive-mind_memory(action: "set", key: "review-economist", value: "<Economist editor's review>")
mcp__ruflo__hive-mind_memory(action: "set", key: "review-lrb", value: "<LRB editor's review>")
mcp__ruflo__hive-mind_memory(action: "set", key: "review-newyorker", value: "<New Yorker editor's review>")

Each review must address:

• Structure: Does the piece have a clear arc? Does it start in the right place?
• Argument: Is the central claim clear, supported, and developed?
• Prose quality: Sentence-level problems — flabbiness, cliche, ambiguity, rhythm
• Emdash audit: Flag every emdash with a specific replacement recommendation
• Cuts: What should be removed? (Be specific — quote the text to cut)
• Additions: What's missing? Where does the piece assert without evidence?
• Headline/lede: Does the opening work? Suggest an alternative if not.

Phase 4: Editorial Debate

Broadcast all four reviews to the hive-mind so each editor can see the others' perspectives:

mcp__ruflo__hive-mind_broadcast(
  message: "All reviews are in. Each editor should now read the other three reviews and prepare their positions for debate.",
  priority: "high"
)

Simulate a structured editorial debate. The editors should:

1. Identify agreement — Where do all four editors converge? These become firm recommendations.
2. Surface disagreements — Where do they diverge? Each editor argues their position.
3. Articulate tradeoffs — Some disagreements reflect genuine editorial tradeoffs (e.g., LRB's tolerance for digression vs The Economist's ruthless compression). Name these tradeoffs explicitly rather than forcing false consensus.
4. Prioritise — Rank all recommendations by impact. What are the three changes that would most improve this piece?

Use the consensus mechanism for firm recommendations:

mcp__ruflo__hive-mind_consensus(action: "propose", type: "editorial-recommendation", value: "<recommendation>")

Each editor votes. Recommendations with unanimous support become "consensus recommendations." Those with majority support become "majority recommendations." Dissents are recorded with reasoning.

Store the debate output:

mcp__ruflo__hive-mind_memory(action: "set", key: "debate-consensus", value: "<consensus recommendations>")
mcp__ruflo__hive-mind_memory(action: "set", key: "debate-tradeoffs", value: "<articulated tradeoffs>")
mcp__ruflo__hive-mind_memory(action: "set", key: "debate-priorities", value: "<top 3 highest-impact changes>")

Phase 5: Present to Author

Present the editorial findings to the user in this structure:

# Masthead Editorial Review

**Document:** <name>
**Genre:** <genre> | **Audience:** <audience> | **Word count:** <approx>

## Top 3 Priorities
<The three highest-impact changes, ranked>

## Consensus Recommendations
<Changes all four editors agree on — these are strong signals>

## Majority Recommendations
<Changes 3 of 4 editors support, with the dissenting view noted>

## Editorial Tradeoffs
<Genuine tensions where the right answer depends on what kind of piece this wants to be>

## Emdash Report
<Every emdash in the document, with the recommended replacement>

## Per-Editor Notes
<Any distinctive insight from a single editor that didn't surface in debate but is worth considering>

Stop here and wait for the user's feedback. Ask which recommendations to accept, which to reject, and whether they have a preference for the target voice.

Phase 6: Incorporate Feedback and Rewrite

Based on the user's direction:

1. Apply all accepted recommendations
2. Rewrite the document in the agreed voice (or the voice the editors recommended)
3. Eliminate all emdashes (unless the user explicitly asks to keep specific ones)
4. Preserve the author's meaning throughout — improve expression, never alter intent
5. Write the revised document back to the original file path

After writing, present a brief summary of changes made.

Phase 7: Shutdown

mcp__ruflo__hive-mind_shutdown(graceful: true)

Important Notes

• Author's intent is sacred. The editors improve how something is said, never what is said. If a recommendation would change the meaning, flag it as a suggestion rather than applying it.
• Emdash policy is strict. The default is elimination. The only exception is interruption in quoted speech. If the user's document uses emdashes as a stylistic choice and they push back on removal, respect their preference — but the editors should still flag them.
• Voice consistency. When rewriting, maintain a consistent voice throughout. Don't let different editors' influences create a patchwork.
• Length. The rewrite should generally be shorter than the original. Good editing compresses.
• Track the debate. The tradeoffs section is as valuable as the recommendations. It helps the author understand the space of editorial choices available to them.