Writers Pattern logo

Writers Pattern

OrganizationPopular
caliber-ai-org
writers-pattern

Add a new platform writer module in src/writers/ that generates and writes agent config files for a supported platform. Each writer exports a function that accepts a config interface, creates directories (rules/, skills/, mcp configs), writes files with proper formatting and frontmatter, and returns string[] of written file paths. Use when adding platform support for a new agent, integrating a new code AI tool, or extending caliber to support new targets. Do NOT use for modifying existing writers, refactoring scoring logic, or changing how writers are invoked.

Overview

Publishercaliber-ai-org
Repositoryai-setup
Skill namewriters-pattern
Stars
1.3K
Forks
124
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by caliber-ai-org on GitHub. Read the source before you install it.

Installation

Install the Writers Pattern AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/caliber-ai-org/ai-setup.git /tmp/ai-setup
mkdir -p .claude/skills
cp -r /tmp/ai-setup/skills/writers-pattern .claude/skills/writers-pattern
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Writers Pattern in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Writers Pattern on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Writers Pattern is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Platform Writer Pattern

Critical

  • Every writer MUST:

    1. Export a single named function: write<Platform>Config(config: <PlatformConfig>): string[]
    2. Accept a platform-specific config interface defining what content to write
    3. Return string[] of all written file paths (used by manifest and progress display)
    4. Create parent directories with fs.mkdirSync(dir, { recursive: true }) before writing files
    5. Wrap skill frontmatter exactly as shown in Step 4 (YAML between --- markers)
    6. NOT modify files outside the intended platform directories (e.g., Claude writer only touches .claude/, CLAUDE.md, .mcp.json)
  • Integration point: Every writer MUST be imported and called in src/writers/index.ts within the writeSetup() function. Missing this step means the writer will never execute.

  • Validation before Step 1: Verify the platform name is unique (ls src/writers/ shows no <platform>/index.ts). If it exists, you are modifying, not adding.

Instructions

Step 1: Define the Platform Config Interface

Verify the platform name is unique (see Critical). Create src/writers/<platform>/index.ts.

At the top of the file, define a config interface that describes all content the writer accepts. The interface MUST include:

  • A main markdown/text file (e.g., claudeMd, cursorrules, agentsMd, instructions)
  • Optional nested arrays: rules, skills, mcpServers, instructionFiles, etc.
  • Each rule/skill/file has at minimum: name or filename, content, and for skills, description

Example (matching GitHub Copilot pattern already in codebase):

typescript
interface CopilotConfig {
  instructions: string;  // Main content
  instructionFiles?: Array<{ filename: string; content: string }>;
}

Validation: Confirm the interface property names match platform conventions (e.g., Claude uses claudeMd, Cursor uses cursorrules). Check existing writers: src/writers/claude/index.ts line 9, src/writers/cursor/index.ts line 9, src/writers/codex/index.ts line 5.

Step 2: Implement the Writer Export Function

Export a function named write<Platform>Config(config: <PlatformConfig>): string[] that:

  1. Initialize an empty written: string[] = [] array to track all written paths.

  2. For the main config file (e.g., CLAUDE.md for Claude, .cursorrules for Cursor):

    • Wrap the content with platform-specific blocks: Use helpers from src/writers/pre-commit-block.ts
    • Common blocks: appendPreCommitBlock(content, platform), appendLearningsBlock(content), appendSyncBlock(content)
    • Examples from actual codebase:
      • Claude (src/writers/claude/index.ts line 19-22): appendSyncBlock(appendLearningsBlock(appendPreCommitBlock(config.claudeMd)))
      • Cursor (src/writers/cursor/index.ts line 19-22): No sync block; injects rules instead (pre-commit, learnings, sync as separate files)
      • Codex (src/writers/codex/index.ts line 13-16): appendLearningsBlock(appendPreCommitBlock(config.agentsMd, 'codex'))
      • Copilot (src/writers/github-copilot/index.ts line 19-22): appendSyncBlock(appendLearningsBlock(appendPreCommitBlock(config.instructions, 'copilot')))
    • Write to the exact path (e.g., fs.writeFileSync('CLAUDE.md', wrappedContent)) and push to written.
  3. For rules (if applicable): Platform convention determines directory.

    • Cursor uses .cursor/rules/, Claude uses .claude/rules/ (see src/writers/index.ts lines 123-126 and 135-137)
    • For each rule: create the directory, write to <dir>/<rule.filename>, push path to written
    • Cursor special case (src/writers/cursor/index.ts line 24-27): Cursor injects three system rules:
      typescript
      const preCommitRule = getCursorPreCommitRule();
      const learningsRule = getCursorLearningsRule();
      const syncRule = getCursorSyncRule();
      const allRules = [...(config.rules || []), preCommitRule, learningsRule, syncRule];
  4. For skills: Write with YAML frontmatter.

    • Directory pattern: .claude/skills/<skillName>/SKILL.md, .cursor/skills/<skillName>/SKILL.md, .agents/skills/<skillName>/SKILL.md, .opencode/skills/<skillName>/SKILL.md
    • Frontmatter format (exact indentation from src/writers/claude/index.ts line 40-47):
      ---
      name: <skill.name>
      description: <skill.description>
      paths:
        - <optional path pattern 1>
        - <optional path pattern 2>
      ---
      <skill.content>
    • For Opencode (src/writers/opencode/index.ts line 30): Use buildSkillContent(skill) from src/lib/builtin-skills.js instead of manual frontmatter.
    • Create directory with fs.mkdirSync(skillDir, { recursive: true }), write skill, push to written
  5. For MCP Servers (if applicable): Write/merge JSON at platform-specific path.

    • Claude (src/writers/claude/index.ts line 54-65): .mcp.json at root
    • Cursor (src/writers/cursor/index.ts line 54-69): .cursor/mcp.json
    • Pattern: Read existing servers (if file exists, try to parse JSON with try/catch), merge with new servers using spread operator { ...existingServers, ...config.mcpServers }, write merged object
    • Wrap in { mcpServers: mergedServers } and output as pretty-printed JSON: JSON.stringify(wrapped, null, 2)
  6. For instruction files (GitHub Copilot, src/writers/github-copilot/index.ts line 26-33): Write to .github/instructions/ directory.

    • Create directory, iterate files, write each to <dir>/<file.filename>, push paths

Return the written array.

Validation: Confirm all file write operations are synchronous (fs.writeFileSync, fs.mkdirSync). Ensure every written path is added to the array. No async operations allowed.

Step 3: Add Type Exports (if complex interface)

If the config interface may be reused elsewhere (e.g., in src/writers/index.ts for the AgentSetup type), export the interface from the module.

Validation: Check src/writers/index.ts lines 15-19 to see if new agent setup params are needed.

Step 4: Import and Register in src/writers/index.ts

Open src/writers/index.ts. At the top (around line 2-6), add:

typescript
import { write<Platform>Config } from './<platform>/index.js';

Update the AgentSetup interface (around line 12-20):

  • Add '<platform>' to the targetAgent tuple (line 13: ('claude' | 'cursor' | 'codex' | 'opencode' | 'github-copilot' | '<platform>')[])
  • Add a new property: <platform>?: Parameters<typeof write<Platform>Config>[0];

Update getFilesToWrite() function (starting line 117): Add a new conditional block:

typescript
if (setup.targetAgent.includes('<platform>') && setup.<platform>) {
  files.push('<main-config-file>');
  if (setup.<platform>.rules) {
    for (const r of setup.<platform>.rules) files.push(`<rules-dir>/${r.filename}`);
  }
  if (setup.<platform>.skills) {
    for (const s of setup.<platform>.skills) files.push(`<skills-dir>/${s.name}/SKILL.md`);
  }
  // ... repeat for other config types (mcpServers, instructionFiles, etc.)
}

Update writeSetup() function (starting line 22): Add a new conditional block before the return (after line 56):

typescript
if (setup.targetAgent.includes('<platform>') && setup.<platform>) {
  written.push(...write<Platform>Config(setup.<platform>));
}

Validation: Confirm the function call order in writeSetup() is consistent (line 37-56): claude → cursor → codex → opencode → github-copilot → (new platform). This ensures AGENTS.md is written once if shared (as with Codex/Opencode, see line 50-52).

Step 5: Add Tests

Create src/writers/__tests__/<platform>.test.ts following the vitest pattern in src/writers/__tests__/codex.test.ts:

  1. Mock fs module: vi.mock('fs')
  2. Mock return values in beforeEach: vi.mocked(fs.existsSync).mockReturnValue(false)
  3. Test that:
    • Main config file is written to correct path
    • Returned array includes all written paths
    • Directories are created before file writes (use .toHaveBeenCalledWith(path, { recursive: true }))
    • Skills have correct frontmatter format (check vi.mocked(fs.writeFileSync).mock.calls)
    • MCP/instruction files are merged/created correctly
    • Pre-commit/learnings/sync blocks are included in main file (expect content .toContain('caliber:managed:pre-commit'))

Run: npm test -- src/writers/__tests__/<platform>.test.ts

Validation: All tests pass. Confirm mocked file operations match actual file system structure.

Step 6: Update detectSyncedAgents() in src/commands/refresh.ts (Optional)

If the platform writes config files with distinct naming (e.g., .newplatform/), update the detection logic around line 68-77 so end-user refresh output correctly identifies the synced platform:

typescript
if (joined.includes('.newplatform/') || joined.includes('newplatform-config')) {
  agents.push('<Platform Name>');
}

Validation: Run npm run refresh and confirm the summary message lists the new platform.

Examples

Example 1: Add a hypothetical "DevCode" platform writer

User says: "Add support for DevCode, a new agent that reads config from .devcode/settings.md and .devcode/rules/ directory."

Actions taken:

  1. Create src/writers/devcode/index.ts (matching pattern from src/writers/claude/index.ts):

    typescript
    import fs from 'fs';
    import path from 'path';
    import { appendPreCommitBlock, appendLearningsBlock } from '../pre-commit-block.js';
    
    interface DevcodeConfig {
      settingsMd: string;
      rules?: Array<{ filename: string; content: string }>;
    }
    
    export function writeDevcodeConfig(config: DevcodeConfig): string[] {
      const written: string[] = [];
    
      fs.writeFileSync(
        '.devcode/settings.md',
        appendLearningsBlock(appendPreCommitBlock(config.settingsMd, 'devcode'))
      );
      written.push('.devcode/settings.md');
    
      if (config.rules?.length) {
        const rulesDir = path.join('.devcode', 'rules');
        if (!fs.existsSync(rulesDir)) fs.mkdirSync(rulesDir, { recursive: true });
        for (const rule of config.rules) {
          const rulePath = path.join(rulesDir, rule.filename);
          fs.writeFileSync(rulePath, rule.content);
          written.push(rulePath);
        }
      }
    
      return written;
    }
  2. Update src/writers/index.ts:

    • Line 2: Add import { writeDevcodeConfig } from './devcode/index.js';
    • Line 13: Change targetAgent tuple to include 'devcode'
    • Line 19: Add devcode?: Parameters<typeof writeDevcodeConfig>[0];
    • Line 117+: Add devcode block to getFilesToWrite() (match Codex pattern lines 144-149)
    • Line 37+: Add devcode block to writeSetup() (match Codex pattern lines 45-47)
    • Line 68+: Update detectSyncedAgents() to check for .devcode/
  3. Create src/writers/__tests__/devcode.test.ts (matching src/writers/__tests__/codex.test.ts):

    typescript
    import { describe, it, expect, vi, beforeEach } from 'vitest';
    import fs from 'fs';
    import path from 'path';
    
    vi.mock('fs');
    
    import { writeDevcodeConfig } from '../devcode/index.js';
    
    describe('writeDevcodeConfig', () => {
      beforeEach(() => {
        vi.clearAllMocks();
        vi.mocked(fs.existsSync).mockReturnValue(false);
      });
    
      it('writes settings.md and rules', () => {
        const config = {
          settingsMd: '# DevCode Config',
          rules: [{ filename: 'style.md', content: 'Style rules' }],
        };
        const written = writeDevcodeConfig(config);
        expect(written).toContain('.devcode/settings.md');
        expect(written).toContain(path.join('.devcode', 'rules', 'style.md'));
      });
    });
  4. Run: npm test && npm run build

Result: Caliber now generates .devcode/settings.md and rules on caliber refresh and caliber init.

Common Issues

Issue: "TypeError: writeConfig is not a function"

  • Fix: Verify the function is exported (not just defined). Check export function write<Platform>Config(...) in the writer file. Missing export is a common mistake.

Issue: "ENOENT: no such file or directory, open '.platform/config.md'"

  • Fix: The parent directory was not created. Ensure fs.mkdirSync(parentDir, { recursive: true }) is called before fs.writeFileSync(filePath, content). See correct order in src/writers/claude/index.ts lines 26-27.

Issue: "Skill file has no frontmatter / malformed YAML"

  • Fix: Verify frontmatter format is exactly (no extra blank lines):
    ---\nname: <name>\ndescription: <desc>\n---\n<content>
    Use [...].join('\n') and test with a single skill first. Compare with working code in src/writers/claude/index.ts lines 40-48.

Issue: "MCP servers not merging, file is truncated"

  • Fix: Confirm the merge pattern from src/writers/claude/index.ts line 54-65: read existing JSON (with try/catch), parse safely, merge with spread operator { ...existingServers, ...config.mcpServers }, then write the merged object. Do NOT overwrite — always merge.

Issue: "new writer is called but written files are empty array"

  • Fix: Verify the writer function returns the written array. Check that every file operation pushes to written. Missing a written.push(filePath) after fs.writeFileSync() is the most common error. See src/writers/codex/index.ts lines 17 and 32 for correct pattern.

Issue: "Tests mock fs but actual files are created in .tmp/ or cause permission errors"

  • Fix: Ensure vi.mock('fs') is at the top of the test file before any imports. All fs operations will be mocked and return mock values from beforeEach setup. See src/writers/__tests__/codex.test.ts line 5.

Issue: "Cursor pre-commit rule not applied, or learnings block missing"

  • Fix: Cursor injects system rules during the write, unlike Claude which uses block appenders. Check that getCursorPreCommitRule(), getCursorLearningsRule(), and getCursorSyncRule() are called and concatenated with user rules before iterating (see src/writers/cursor/index.ts lines 24-27). Claude and Codex use block-append helpers instead.

Frequently asked questions

What does the Writers Pattern AI skill do?

Add a new platform writer module in src/writers/ that generates and writes agent config files for a supported platform. Each writer exports a function that accepts a config interface, creates directories (rules/, skills/, mcp configs), writes files with proper formatting and frontmatter, and returns string[] of written file paths. Use when adding platform support for a new agent, integrating a new code AI tool, or extending caliber to support new targets. Do NOT use for modifying existing writers, refactoring scoring logic, or changing how writers are invoked.

Why use Writers Pattern on TypingMind?

Because you install it once and use it with any model. Writers Pattern is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Writers Pattern in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/caliber-ai-org/ai-setup/tree/master/skills/writers-pattern. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Writers Pattern?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Writers Pattern?

As many as you like. As long as a model supports skills, you can use Writers Pattern with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Writers Pattern AI skill free?

Yes. It is published on GitHub by caliber-ai-org under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇