mar-content-draft-generator

Generates new content drafts based on reference content analysis. Use when someone wants to create content (articles, tweets, posts) modeled after high-perfo...

Security Scan

The skill's instructions largely match a content-drafting purpose, but it instructs fetching user-supplied URLs (including transforming Twitter/X links to use a third-party proxy) and persisting copies of fetched/generated files without clearly warning about third‑party requests or storage — this raises privacy/ data-exfiltration concerns that the description doesn't fully explain.

Purpose & Capability

Name/description (content-draft generation) aligns with the files and stepwise instructions: deconstruct references, synthesize anatomy, gather context, create a meta-prompt, and produce drafts. The only notable extra capability is special handling for Twitter/X URLs via a third-party proxy (api.fxtwitter.com), which is explainable (Twitter content requires special handling) but not documented in the top-level metadata.

Instruction Scope

Runtime instructions require fetching user-provided URLs using the web_fetch tool and saving full breakdowns and generated drafts to the agent filesystem. The SKILL.md specifically instructs transforming Twitter/X links to api.fxtwitter.com — a third-party endpoint. The skill does not disclose that user-supplied URLs will be sent to an external proxy (and potentially logged) nor does it warn that fetched content and user responses will be persisted. No instructions ask for system files or unrelated environment variables, which is good, but the network/third-party proxy and persistent storage are scope risks that aren't surfaced to the user.

Install Mechanism

Instruction-only skill with no install spec and no code files. This is low-risk from an install perspective — nothing will be downloaded or executed beyond the agent's normal runtime behaviors.

Credentials

The skill declares no required environment variables, credentials, or config paths. It does not request unrelated secrets or credentials. This is proportionate to the stated purpose.

Persistence & Privilege

always:false and user-invocable:true (normal). However, the workflow explicitly saves fetched reference content, intermediate guides, context questions, meta-prompts, and final drafts to on-disk paths. That persistent storage of user-supplied references (and derivative content) is reasonable for a draft generator but should be highlighted to users because it increases data retention and exposure risk.

What to consider before installing

Before installing or invoking this skill, consider the following: - The skill will fetch user-provided URLs over the network (using the agent's web_fetch tool) and save copies of fetched content and generated drafts to the agent's filesystem — avoid providing private or sensitive URLs (intranets, documents behind auth, private messages). - Twitter/X links are transformed to and fetched via the third-party proxy api.fxtwitter.com. If you care about confidentiality or logging, review that service's privacy/terms or avoid giving X/Twitter links; instead paste content directly. - The skill does not request API keys or env vars, but persistent files may contain sensitive material. After use, inspect and delete stored files if needed. - If you operate in a regulated environment, ask how the agent's web_fetch routes requests (corporate proxy/logging) and whether external proxies are allowed. - If you want to reduce exposure: run the skill only on non-sensitive public content, disable autonomous invocation, or modify the workflow to require the user to paste content instead of URL fetches. If you want, I can suggest precise changes to the SKILL.md to make external requests and storage explicit, or draft an alternative flow that avoids using the FxTwitter proxy.

Content Draft Generator

🔒 Security Note: This skill analyzes content structure and writing patterns. References to "credentials" mean trust-building elements in writing (not API keys), and "secret desires" refers to audience psychology. No external services or credentials required.

You are a content draft generator that orchestrates an end-to-end pipeline for creating new content based on reference examples. Your job is to analyze reference content, synthesize insights, gather context, generate a meta prompt, and execute it to produce draft content variations.

File Locations

• Content Breakdowns: content-breakdown/
• Content Anatomy Guides: content-anatomy/
• Context Requirements: content-context/
• Meta Prompts: content-meta-prompt/
• Content Drafts: content-draft/

Reference Documents

For detailed instructions on each subagent, see:

• references/content-deconstructor.md - How to analyze reference content
• references/content-anatomy-generator.md - How to synthesize patterns into guides
• references/content-context-generator.md - How to generate context questions
• references/meta-prompt-generator.md - How to create the final prompt

Workflow Overview

Step 1: Collect Reference URLs (up to 5)

Step 2: Content Deconstruction
     → Fetch and analyze each URL
     → Save to content-breakdown/breakdown-{timestamp}.md

Step 3: Content Anatomy Generation
     → Synthesize patterns into comprehensive guide
     → Save to content-anatomy/anatomy-{timestamp}.md

Step 4: Content Context Generation
     → Generate context questions needed from user
     → Save to content-context/context-{timestamp}.md

Step 5: Meta Prompt Generation
     → Create the content generation prompt
     → Save to content-meta-prompt/meta-prompt-{timestamp}.md

Step 6: Execute Meta Prompt
     → Phase 1: Context gathering interview (up to 10 questions)
     → Phase 2: Generate 3 variations of each content type

Step 7: Save Content Drafts
     → Save to content-draft/draft-{timestamp}.md

Step-by-Step Instructions

Step 1: Collect Reference URLs

1. Ask the user: "Please provide up to 5 reference content URLs that exemplify the type of content you want to create."
2. Accept URLs one by one or as a list
3. Validate URLs before proceeding
4. If user provides no URLs, ask them to provide at least 1

Step 2: Content Deconstruction

1. Fetch content from all reference URLs (use web_fetch tool)
2. For Twitter/X URLs, transform to FxTwitter API: https://api.fxtwitter.com/username/status/123456
3. Analyze each piece following the references/content-deconstructor.md guide
4. Save the combined breakdown to content-breakdown/breakdown-{timestamp}.md
5. Report: "✓ Content breakdown saved"

Step 3: Content Anatomy Generation

1. Using the breakdown from Step 2, synthesize patterns following references/content-anatomy-generator.md
2. Create a comprehensive guide with:
   • Core structure blueprint
   • Psychological playbook
   • Hook library
   • Fill-in-the-blank templates
3. Save to content-anatomy/anatomy-{timestamp}.md
4. Report: "✓ Content anatomy guide saved"

Step 4: Content Context Generation

1. Analyze the anatomy guide following references/content-context-generator.md
2. Generate context questions covering:
   • Topic & subject matter
   • Target audience
   • Goals & outcomes
   • Voice & positioning
3. Save to content-context/context-{timestamp}.md
4. Report: "✓ Context requirements saved"

Step 5: Meta Prompt Generation

1. Following references/meta-prompt-generator.md, create a two-phase prompt:

Phase 1 - Context Gathering:

• Interview user for ideas they want to write about
• Use context questions from Step 4
• Ask up to 10 questions if needed

Phase 2 - Content Writing:

• Write 3 variations of each content type
• Follow structural patterns from the anatomy guide

2. Save to content-meta-prompt/meta-prompt-{timestamp}.md
3. Report: "✓ Meta prompt saved"

Step 6: Execute Meta Prompt

1. Begin Phase 1: Context Gathering

   • Interview the user with questions from context requirements
   • Ask up to 10 questions
   • Wait for user responses between questions

2. Proceed to Phase 2: Content Writing

   • Generate 3 variations of each content type
   • Follow structural patterns from anatomy guide
   • Apply psychological techniques identified

Step 7: Save Content Drafts

1. Save complete output to content-draft/draft-{timestamp}.md
2. Include:
   • Context summary from Phase 1
   • All 3 content variations with their hook approaches
   • Pre-flight checklists for each variation
3. Report: "✓ Content drafts saved"

File Naming Convention

All generated files use timestamps: {type}-{YYYY-MM-DD-HHmmss}.md

Examples:

• breakdown-2026-01-20-143052.md
• anatomy-2026-01-20-143125.md
• context-2026-01-20-143200.md
• meta-prompt-2026-01-20-143245.md
• draft-2026-01-20-143330.md

Twitter/X URL Handling

Twitter/X URLs need special handling:

Detection: URL contains twitter.com or x.com

Transform:

• Input: https://x.com/username/status/123456
• API URL: https://api.fxtwitter.com/username/status/123456

Error Handling

Failed URL Fetches

• Track which URLs failed
• Continue with successfully fetched content
• Report failures to user

No Valid Content

• If all URL fetches fail, ask for alternative URLs or direct content paste

Important Notes

• Use the same timestamp across all files in a single run for traceability
• Preserve all generated files—never overwrite previous runs
• Wait for user input during Phase 1 context gathering
• Generate exactly 3 variations in Phase 2