weekly-plan-review

name	weekly-plan-review
description	Re-visit posted URLs from the previous week's plan and scrape current likes / reach into the Metrics sheet.

Common rules: docs/OPERATING_RULES.md.

Input

• prev_plan_json_path: reports/plans/<brand>/<prev iso_week>/plan.json.

Output

{
  "metrics_added": 12,
  "stats_history_updated": true
}

Steps

1. Load prev_plan_json_path. Iterate Schedule rows with status: posted and a posted_url.

2. For each row, navigate to posted_url with the logged-in Playwright session and read public engagement metrics:

   • facebook: likes/reactions, comments, shares (visible on the post card).
   • instagram: likes (post), views (Reel), comments. Saves are visible only in Insights — skip if not exposed.
   • x: likes, replies, reposts, views.
   • threads: likes, replies, reposts (no view counts as of plan-write).
   • youtube: views (visible on watch page), likes (toggle if hidden).
   • tiktok: views, likes, comments, shares.

3. Append a row to Metrics:

   {
     "slot_id": "<from Schedule>",
     "posted_url": "...",
     "metrics_captured_at": "<ISO>",
     "likes": 123,
     "reach": 4567,
     "comments": 12,
     "saves": null,
     "notes": ""
   }
   

4. Roll up into data/stats-history/<brand>.json:

   {
     "brand": "noirs-boxes",
     "by_platform": {
       "instagram": {
         "totals_by_iso_week": { "2026-W18": { "likes": ..., "reach": ..., "post_count": 4 } },
         "by_weekday": { "Mon": { "mean_likes": ..., "n": ... } },
         "by_time_band": { "20:00-21:00": { "mean_likes": ..., "n": ... } }
       
     
   
   

5. Save the updated plan.json and re-export to plan.xlsx via weekly-plan-export.

6. mcp__playwright__browser_close.

Quirks

• Many platforms hide exact view counts unless logged in as the page owner. The logged-in profile already covers own pages — that's why this is "own posts only".
• Some metrics drift over weeks (likes accrue, view counts climb). Capture the snapshot at the time of review and rely on metrics_captured_at to interpret freshness.

Don'ts

• Do not scrape competitor accounts — they live in research, not metrics.
• Do not write to posted_url rows from a plan that was never published (status != posted).
• Do not call other skills except weekly-plan-export.
• Do not skip browser_close.