Page Path Evaluator

Context: Page Path Evaluator

This internal skill evaluates a captured page snapshot against the expected URL patterns and current exploration goal.

Inputs

Provide the following in your invocation:

1. Snapshot: {url, title, text, links}
2. Goal: The current exploration goal (e.g., "Find API design discussion")
3. Visited Set: List of already visited URLs
4. Current Queue: List of currently queued URLs

Execution Steps

1. Classify Page:

   • Compare url against patterns in skills/internal/context/url-patterns/SKILL.md.
   • Identify page_type.

2. Calculate Confidence:

   • Check text for Forbidden Content for that type (if found, confidence is 0%).
   • Count how many Expected Content keywords appear in text.
   • Score = (found / expected_total) * 100.
   • Adjust score based on relevance to the specific Goal (e.g., if goal mentions "API design" and text contains it, +20%).

3. Extract & Rank Links:

   • For each link in the snapshot:
     • Ignore if in Visited Set or Current Queue.
     • Score relevance: Same-site (+10), Matches goal keywords in anchor text (+30), Matches expected priority extraction pattern for this type (+20).
   • Sort links descending by score.
   • Keep the top 3.

4. Return Output: Format the result as a structured JSON object:

   {
     "type": "github_pr",
     "confidence": 85,
     "summary": "Found PR description with 12 comments discussing API design.",
     "next_links": [
       {"url": "...", "title": "Files changed", "reason": "Standard next step for PR"},
       {"url": "...", "title": "TIS-999", "reason": "Mentioned in description"}
     ]
   }