blog post pipeline

Manual blog production often suffers from hallucinated facts, drifting brand voice, and fragmented workflows. This skill implements a rigorous, gated editorial pipeline that enforces truth and consistency across multiple publications.

What it does

• Enforces veracity by blocking any quantitative or comparative claims that lack a cited source from the research phase.
• Isolates roles to prevent the writer from researching or the reviewer from self-editing, ensuring objective adversarial oversight.
• Manages state via a file-based system, allowing you to pause and resume complex content projects across different sessions.
• Handles localization through cultural adaptation rather than literal translation, keeping the message intact across languages.
• Automates publishing including image planning, asset uploads, and final Markdown deployment to target repositories.

How it works

1. Initialize the brief by creating a configuration file that defines the blog target and primary goals.
2. Execute research and curation where the agent gathers sources and the human selects which facts are approved for the draft.
3. Generate and review the draft, triggering a specialized adversarial pass that flags unsourced claims as blockers.
4. Finalize and publish by optionally translating the content, planning media, and pushing to the target environment.

Frameworks & tools

The skill is designed for environments with filesystem access like Claude Code or Cursor. It uses YAML for brand configuration and maintains state through a structured Markdown file hierarchy.

Why this beats prompting it yourself

Generic prompts often lead to "hallucinated expertise" where the AI invents statistics to sound persuasive. this skill forces a structural separation between fact-finding and prose generation, making it impossible for the agent to include unverified data.

Use cases

• Technical content marketing requiring precise performance benchmarks and cited specifications.
• Bilingual blog management where posts must be adapted for different cultural contexts while maintaining a single source of truth.
• High-volume editorial teams needing a repeatable, audited process for multi-author or multi-brand environments.

Known limitations

Requires initial manual setup of a blogs.yaml file for brand voice and hard limits. Does not support automated chaining of stages; a human must approve each gate to progress the state.