Blog Generator

Analyzes journal entries and chat history to identify high-value topics and automatically generate blog posts.

Security Scan

The skill mostly matches its stated purpose (scanning local journals and generating posts) but contains several incoherencies and potential privacy risks (undeclared external-humanizer/API usage, hardcoded user paths, missing referenced files).

Purpose & Capability

The name/description (generate blog posts from journal/chat history) matches the included scripts which scan ~/.openclaw/journal and write to ~/.openclaw/blogs, so core capabilities align. However the SKILL.md and scripts reference an external 'humanizer' that requires OPENROUTER_API_KEY and a visual-explainer header generator (scripts/generate_header.py) that is not present in the manifest — those external integrations are not declared in metadata (no required env vars), which is inconsistent with the skill's stated, local-only purpose.

Instruction Scope

Runtime instructions and scripts instruct the agent to read a user's journal directory, scan many markdown files, and save HTML to a hardcoded user path (/Users/ghost/.openclaw/...). The SKILL.md also suggests running a humanizer between generations and pulling a header from a separate skill path; both could cause content to be sent outside the machine depending on implementation. The SKILL.md grants broad discretion (cron, scheduled autonomous runs) and references external models/APIs in examples (openrouter/google/gemini-2.5-flash) not declared in requirements.

Install Mechanism

No install spec is provided (instruction-only + included scripts). No network download/install steps are present in the manifest. That limits risk from arbitrary installers. The code does import subprocess and likely invokes local binaries (humanizer, generate_header) — this is expected but those binaries aren't included here.

Credentials

The skill metadata declares no required environment variables, yet SKILL.md and code mention OPENROUTER_API_KEY (for the humanizer) and show example cron payloads using an OpenRouter model. Requiring an API key to call an external service is reasonable only if declared; here the omission is a mismatch. Asking for an API key that would enable sending journal content externally is disproportionate unless explicitly stated and justified.

Persistence & Privilege

always:false (no forced inclusion) and default autonomous invocation are normal. The skill writes files into the user's ~/.openclaw/blogs directory (expected for a blog generator). It does not request system-wide privileges or modify other skills' configs in the provided files. Still, scheduling it as a cron job (recommended in SKILL.md) increases exposure if the humanizer or other integrations transmit data externally.

What to consider before installing

This skill will read your OpenClaw journal files and write generated blog HTML into ~/.openclaw/blogs; that behavior is consistent with its purpose. However: (1) SKILL.md and the scripts reference a 'humanizer' that requires OPENROUTER_API_KEY and call an external model in examples — the skill metadata does not declare any required credentials. If the humanizer or model send data to external services, your private journal content could be transmitted. (2) Several paths are hardcoded to /Users/ghost and a referenced generate_header.py (visual-explainer) is not included — you should update paths before running. (3) README and SKILL.md disagree about output format (.md vs .html). Recommended steps before installing or scheduling: inspect the rest of scripts/blog_generator.py (the file was truncated in the package you received) and specifically find implementations of run_humanizer and generate_header to see whether they perform network calls or subprocess spawns; run the skill locally with --no-humanize to avoid external calls; do not set or export OPENROUTER_API_KEY unless you have reviewed and accept where data will be sent; test the scripts in a sandboxed account or VM; and confirm or replace hardcoded /Users/ghost paths with your own home directory. If you need, ask the author for a manifest update that declares required environment variables and the external components the skill depends on.

Blog Generator | OpenClaw Skill

Description

Analyzes journal entries and chat history to identify high-value topics and automatically generate blog posts.

Blog Generator | OpenClaw Skill

Automatically generates blog posts by analyzing journal entries, chat history, and recent activity to identify high-value, high-search-volume topics related to OpenClaw.

Usage

• As a scheduled cron job to automatically generate blog content weekly or daily
• Manually to create blog posts from recent journal analysis
• To identify and document high-value solutions and discoveries

# X-format articles as HTML; humanizer runs between generations; header from visual-explainer
python3 /Users/ghost/.openclaw/workspace/skills/blog-generator/scripts/blog_generator.py

# Skip humanizer (e.g. no OPENROUTER_API_KEY)
python3 /Users/ghost/.openclaw/workspace/skills/blog-generator/scripts/blog_generator.py --no-humanize

# Custom humanizer or visual-explainer paths
python3 /Users/ghost/.openclaw/workspace/skills/blog-generator/scripts/blog_generator.py --humanizer-path /Users/ghost/Downloads/humanizer-1.0.0 --visual-explainer-path /Users/ghost/.openclaw/workspace/skills/visual-explainer-main

# Classic format (overview/problem/solution), still HTML
python3 /Users/ghost/.openclaw/workspace/skills/blog-generator/scripts/blog_generator.py --format classic

# JSON output
python3 /Users/ghost/.openclaw/workspace/skills/blog-generator/scripts/blog_generator.py --days 14 --max-topics 5 --json

What this skill does

• Scans journal entries from the last N days for interesting topics (discoveries, obstacles, solutions)
• Identifies high-value topics based on keyword relevance and problem-solving value
• Researches search volume and keyword opportunities (heuristic-based, can be enhanced with APIs)
• Generates structured blog posts with overview, problem, solution, and takeaways sections
• Saves blog posts to /Users/ghost/.openclaw/blogs/ as HTML only (X-article format, header from visual-explainer, humanizer between generations)

Integration as a Cron Job

This skill is designed to run periodically (daily or weekly) via OpenClaw cron to automatically generate blog content.

Example Cron Job Configuration (Daily):

{
  "payload": {
    "kind": "agentTurn",
    "message": "Run blog-generator skill to analyze journal entries and generate high-value blog posts.",
    "model": "openrouter/google/gemini-2.5-flash",
    "thinking": "low",
    "timeoutSeconds": 300
  },
  "schedule": {
    "kind": "cron",
    "cron": "0 9 * * *"
  },
  "delivery": {
    "mode": "announce"
  },
  "sessionTarget": "isolated",
  "name": "Blog Post Generator"
}

Example Cron Job Configuration (Weekly):

{
  "payload": {
    "kind": "agentTurn",
    "message": "Run blog-generator skill with --days 7 --max-topics 3 to generate weekly blog posts from journal analysis.",
    "model": "openrouter/google/gemini-2.5-flash",
    "thinking": "low",
    "timeoutSeconds": 300
  },
  "schedule": {
    "kind": "cron",
    "cron": "0 10 * * 1"
  },
  "delivery": {
    "mode": "announce"
  },
  "sessionTarget": "isolated",
  "name": "Weekly Blog Generator"
}

Output Format

HTML only (no Markdown). Output is in the format accepted by X articles: one header (from visual-explainer), then article body with no interlaced visuals.

• File: /Users/ghost/.openclaw/blogs/YYYYMMDD_slugified-title.html
• Article structure: X-style (long-form): punchy hook, short paragraphs, "two types of people" framing, pivot, stakes. Use --format classic for overview/problem/solution.
• Header: From visual-explainer only. Default path: /Users/ghost/.openclaw/workspace/skills/visual-explainer-main. The skill’s scripts/generate_header.py is called with section: "header" and returns an html_snippet (hero with title and optional summary). No diagrams or images in the body.
• Humanizer: Runs between generations by default. Each article’s body is sent through /Users/ghost/Downloads/humanizer-1.0.0 (requires OPENROUTER_API_KEY) before rendering to HTML. Use --no-humanize to skip.

Topic Scoring

Topics are scored based on:

• High-value keywords: OpenClaw-specific terms, problem-solving language
• Content type: Solutions score highest, then obstacles, then discoveries
• Content depth: Longer, more detailed content scores higher
• Search volume indicators: Keywords like "how to", "tutorial", "fix" increase value

Requirements

• Journal entries in /Users/ghost/.openclaw/journal/
• Blogs directory writable at /Users/ghost/.openclaw/blogs/
• Chat history analyzer skill (for journal entries)

How it works

1. Scans journal directory for markdown files from the last N days
2. Extracts topics from discoveries, obstacles, and solutions sections
3. Scores topics based on keyword relevance and value
4. Selects top N high-value topics
5. Generates structured blog posts with problem/solution format
6. Saves posts to blogs directory with timestamped filenames

Enhancement Opportunities

• Integrate with Google Keyword Planner API for real search volume data
• Use AI model to enhance blog post quality and SEO optimization
• Cross-reference with existing blog posts to avoid duplicates
• Generate multiple variations of posts for A/B testing