Run2 pdf calendar parsing

Skill cxcscmu/SkillLearnBench/skills/b2-self-feedback-claude-sonnet-4-6/schedule-planning/run2_pdf-calendar-parsing

Precisely parse a PDF day-view calendar to extract event times, durations, and colors by reading visual block positions and the 15-minute grid.From its SKILL.md

Install

npx -y skills add cxcscmu/SkillLearnBench --skill run2_pdf-calendar-parsing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

2.2 KB, 605 tokens by cl100k_base, as published. Nobody here has run it

PDF Calendar Parsing (Improved)

Grid Structure

Each hour label (12am, 1am, …, 11pm) marks the TOP of that hour's row
Between two adjacent dashed horizontal lines = 15 minutes
4 intervals per hour

Reading Block Times — Step by Step

Find the hour label closest ABOVE the block's top edge
Count how many 15-min ticks down from that label to the block's top → start time
Find the hour label closest ABOVE the block's bottom edge
Count ticks to block bottom → end time
If a block starts exactly at a label, start = that label's time on the dot

Color Classification

Color	Hex range (approx)	Status
Dark gray	R<100, G<100, B<100	Busy (OOO)
Red/salmon	R>200, G<150, B<150	Busy
Blue/purple	R~~130, G~~130, B~200-230	Available (overwritable)
Lime/yellow-green	R>150, G>180, B<100	Busy

Example Calendar — March 9, 2026

12:00 AM – 08:00 AM  Out of office         (gray)   BUSY
08:00 AM – 10:00 AM  [empty]                         FREE
10:00 AM – 11:00 AM  Weekly Group Meeting  (red)    BUSY
11:00 AM – 12:00 PM  Busy with task B      (BLUE)   FREE (overwritable)
12:00 PM – 02:00 PM  [empty]                         FREE
02:00 PM – 03:00 PM  Project A Discussion  (red)    BUSY
03:00 PM – 03:15 PM  Coffee Chat           (lime)   BUSY  ← 1 tick = 15 min
03:15 PM – 06:00 PM  [empty]                         FREE
06:00 PM – 12:00 AM  Out of office         (gray)   BUSY

Thin Block Disambiguation

A block taking exactly ONE 15-min slot looks like a thin line just above a dashed separator.

If the label is at 3pm and the block bottom aligns with the first dashed line below → 3:00–3:15

Python extraction with pdfplumber

import pdfplumber

with pdfplumber.open("calendar.pdf") as pdf:
    text = pdf.pages[0].extract_text()
    # Text output lists event names and time labels
    # Visual block colors require image analysis

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.