/check

Context

Read the skill definition for output formats and classification criteria:

• @skills/distill/SKILL.md
• @skills/distill/references/classification.md

Task Workflow

URL: $1

Step 0: Find Vault

0. Look for .distill-state.yaml in the current directory, then ~/Obsidian/Insights/. If not found, ask the user for the vault path.

Step 1: Fetch

1. WebFetch the URL. If it fails, report the error and ask if the user wants to classify based on title alone. (WebFetch is fine for single URLs - the curl approach is only needed for batch scraping.)

Step 2: Classify

2. Read the content and apply the insight heuristic from classification.md.
3. Present the verdict to the user:

If insight:

Verdict: insight ({confidence})
Summary: {one-line insight summary}
Concepts: {tags}
Topic: {proposed folder}

If noise:

Verdict: noise
Reason: {why it doesn't meet the insight bar}

If unclear:

Verdict: borderline
For: {what makes it potentially insightful}
Against: {what makes it potentially noise}

4. Ask the user: "Add to vault?" (for insight/borderline) or "Add to noise log?" (for noise).

Step 3: Write

5. If user says yes to insight:

   • Read .distill-state.yaml for vault path
   • Generate full insight note (frontmatter + Key Insight + Takeaways + Context + scraped content)
   • Write to {vault}/{topic}/{filename}.md

6. If user says yes to noise:

   • Append to {vault}/_noise.md: - [Title](url) — noise: {reason}

7. If user says ignore: do nothing.

Expected Outcome

• Quick classification of a single URL with reasoning shown
• User decides whether to add
• Note or noise log entry written if approved