| name | draft-ingestion |
| description | Ingests raw text, threads, or notes from the inbox/ directory, refines the tone, and formats them into standard markdown posts or notes. |
Draft Ingestion
This skill defines the workflow for turning raw text files (like Perplexity threads, raw ideas, or markdown dumps) in the inbox/ directory into published content.
Workflow
- Check the Manifest:
- Before processing, consult
inbox/manifest.yaml.
- If a file is not in the manifest, run the
manifest-update skill first.
- Files in
inbox/_ignore/ are automatically skipped and not tracked.
- Use the
description and suggested_target from the manifest to guide the ingestion.
- Read the Raw Dump:
- Locate the raw text file in the
inbox/ directory. Read its contents.
- Apply Persona (Human Writer):
- Refine the text using the principles from the
human-writer skill to ensure the tone matches the site's authentic, direct style. Strip out overly enthusiastic AI tropes, robotic transitions, and unnecessary boilerplate.
- Structure and Format:
- Decide if the content is a short-form
note (goes in src/notes/) or a long-form post (goes in src/posts/).
- Add standard frontmatter (
title, description, date, tags, authored_by).
- If third-party attribution is needed, use
author, source, authorUrl, sourceUrl frontmatter instead of hardcoded HTML callouts.
- Cleanup:
- Delete the raw file from the
inbox/ directory once successfully ingested.
- Assets:
- If the new content would benefit from an illustration, prompt the user if you should run the
blog-sketches skill.