🧪 [EXPERIMENTAL] Add compound-writing plugin for AI-powered writing tools #97
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Create a new standalone plugin that applies compound engineering principles to content creation. Just as compound-engineering captures coding patterns, compound-writing captures voice, style, and editorial preferences to accelerate writing. Components: - 7 specialized agents (source-researcher, fact-checker, structure-architect, voice-guardian, clarity-editor, publishing-optimizer, every-style-editor) - 4 workflow commands (writing:plan, writing:draft, writing:review, writing:compound) - 5 skills (pragmatic-writing, dhh-writing, voice-capture, every-style-editor, writing-orchestration) Key features: - Two-agent orchestration pattern (Orchestrator + Writer) - Two-gate content readiness assessment - 10 baseline writing strategies (always apply) - 20+ situational strategies (select 3-4) - Producer-critic loop with voice score targeting - Quality checkpoints for draft verification - Pattern capture system for compounding knowledge Moved every-style-editor from compound-engineering to compound-writing as it's more appropriate for writing workflows. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…systems New commands: - /writing:feedback - Capture real-time preferences into scratchpad - /writing:knowledge - Manage voice profiles, patterns, and references New skills: - scratchpad - Session memory for iterative refinement - context-notes - Structured handoff protocol between commands Updates: - All commands now support context notes for seamless handoffs - Draft ID management for persistent tracking across session - Writing-orchestration skill integrated with scratchpad - README updated with comprehensive workflow documentation Component counts: 7 agents, 6 commands, 7 skills 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
- WRITING-ARCHITECTURE.md: Complete 8-phase writing process framework - 18 agent roles mapped across phases - 5 skill types (core, voice, style guide, domain, format) - "10 books expert" pattern for domain expertise - Implementation roadmap - CLAUDE.md: Development guide for plugin contributors - Current component inventory - Key concepts (two-agent, two-gate, strategies) - Adding new agents/skills/voices - Quality checklist 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…skill loading - Add "Skills to Load" sections to all 6 commands - Number all steps explicitly (Step 1, Step 2, etc.) - Add AskUserQuestion blocks for interactive decision points - Mark brainstorming steps with (BRAINSTORM) labels - Add quality checklists at the end of each command - Include context notes handoff sections Commands enhanced: - /writing:plan (11 steps) - /writing:draft (14 steps) - /writing:review (15 steps) - /writing:compound (12 steps) - /writing:feedback (8 steps) - /writing:knowledge (multiple actions with steps) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
New skill capturing Every's essay philosophy and best practices: - Worldview: optimistic realism, personal narrative + intellectual rigor - 10 craft principles (hook, depth, facts before feelings, etc.) - Hed/Dek/Opening paragraph playbook - Pre-share checklist (8 quality scans) - Quick reference card Complements every-style-editor which handles mechanical editing. Total skills: 7 → 8 Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
New skill for Every's Source Code series - essays on building in the AI era: - Vision: Coin language, demonstrate workflows, shape industry thinking - Source Code gradient (engineering → practice → philosophy → journey) - 6 strategy principles (first-person authority, stakes, One Big Idea, ELI5) - Sensibility and voice spectrum - Opening patterns and structural templates - 10 Source Code Commandments Total skills: 8 → 9 Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Restructure skills directory following WRITING-ARCHITECTURE.md types: skills/ ├── core/ # Orchestration, memory, handoffs (4) ├── voice/ # All voices - can layer multiple (4) │ ├── pragmatic-writing │ ├── dhh-writing │ ├── every-essay-guide │ └── source-code-guide ├── editing/ # Review tools for post-draft (1) │ └── every-style-editor ├── domain/ # Future: "10 Books Expert" skills └── format/ # Future: platform-specific Key insight: every-essay-guide and source-code-guide are VOICE skills (they define how to write for a publication), while every-style-editor is an EDITING tool (checks mechanical compliance after drafting). Voice layering now supported - combine multiple voices per piece. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
New command to create voice, domain, and format skills automatically: Usage: - /writing:skill create voice "kieran-blog" samples/*.md - /writing:skill create domain "saas-metrics" - /writing:skill create format "linkedin" Features: - Parallel sub-agents research the skill type - Reads existing skills as patterns/templates - Generates complete skill structure with references - Updates documentation automatically - Commits the new skill Total commands: 6 → 7 Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Update skill templates with concrete writing examples: Voice skill: - Before/after paragraph transformations - Writing-specific anti-patterns (passive voice, jargon, hedging) Domain skill: - "Use when writing" context - Writing frameworks and argument structures - Example paragraphs with hooks and key phrases Format skill: - Full example post with annotation - Hook examples (contrarian, story, number, bold claim) - Writing techniques (front-load value, write for skimmers) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…ject setup Interactive wizard that sets up compound-writing for a project: - Creates .claude/writing-knowledge/ directories - Initializes drafts/ with scratchpad and context files - Creates/updates project CLAUDE.md with writing config - Configures voice profile (from samples, description, or preset) - Sets default style guide preference - Optional .gitignore configuration Supports --minimal mode for quick setup without questions. Co-Authored-By: Claude <noreply@anthropic.com>
…ptions - Add **MANDATORY** instruction to ensure tool is used, not plain text - Convert all "What next?" sections to explicit YAML format - Add "Based on selection" handlers for each option - Reorder options to put "Open/View" first (review before action) - Add "Open outline/draft/report" as primary action Updated commands: - writing-plan: Open outline → Deepen → Review → Start drafting - writing-draft: Open draft → Review → Feedback → Refine - writing-review: View report → View diff → Another pass → Compound - writing-compound: View patterns → Review library → Start new - writing-skill: View skill → Test → Refine → Create another - writing-init: View setup → Start draft → Configure → Done Co-Authored-By: Claude <noreply@anthropic.com>
- Add "First Time Setup" section with /writing:init - Add "Create Custom Skills" section with /writing:skill examples - Reorganize Quick Start into logical sections - Show both interactive and --minimal setup modes Co-Authored-By: Claude <noreply@anthropic.com>
Add automated prose linting using Vale with 15 custom rules based on
Every's style guide. Rules are bundled inside the skill and called via
${CLAUDE_PLUGIN_ROOT}/skills/editing/every-style-editor/scripts/lint.sh
Vale rules included:
- OverusedWords (actually, very, just)
- PassiveVoice detection
- EmDash spacing (no spaces)
- OxfordComma enforcement
- ClickHere link text
- StartWithThis sentence starts
- WeHaveWeGet openings
- Cliches and jargon
- Percentages format
- CompanyPronouns (it not they)
- BoldEmphasis (use italics)
- Numbers (spell out 1-9)
- Adverbs (cuttable adverbs)
- Ellipsis spacing
- OverUnder (more/fewer than)
Updated /writing:review to run Vale via skill script.
Updated /writing:skill to optionally create Vale rules for new skills.
Sources:
- https://vale.sh/docs/
- https://vale.sh/docs/topics/styles/
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude <noreply@anthropic.com>
- Add ai-pattern-detector agent for detecting AI slop - Add Vale rules: AISlop.yml (flagged words) and AIPhrases.yml (formulaic phrases) - Update /writing:plan to ASK about research instead of auto-searching - Bump version to 1.2.0 (now 8 agents, 8 commands, 9 skills) Sources: - https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing - https://aisdr.com/blog/words-to-avoid-so-you-dont-sound-like-ai/ Co-Authored-By: Claude <noreply@anthropic.com>
- /writing:plan now searches YOUR library first (always) - Past pieces in drafts/ and published/ - Patterns in .claude/writing-knowledge/patterns/ - Voice profiles in .claude/writing-knowledge/voices/ - Scratchpad preferences - Web research is now OPTIONAL (asked after library search) - /writing:init creates proper library structure: - voices/, patterns/, pieces/, references/ - published/ archive for finished work Your past work compounds. External research is supplementary. Co-Authored-By: Claude <noreply@anthropic.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Adds a new compound-writing plugin focused on AI-powered writing workflows. This is a sister plugin to compound-engineering, applying the same "compounding" philosophy to writing tasks.
What's included (EXPERIMENTAL)
/writing:plan,/writing:draft,/writing:edit,/writing:publish)Key Features (EXPERIMENTAL)
Installation (EXPERIMENTAL)
This plugin is:
Use at your own risk. Feedback welcome!
🤖 Generated with Claude Code