screenshot-to-linear

Screenshot to Linear

Upload screenshots from iOS Simulator MCP or Playwright MCP to Linear as inline attachments. Handles the full pipeline: capture → resize → encode → upload → embed.

When to Use

• During the autosymph verify state (iOS or web)
• When the verify agent needs to upload evidence to Linear
• When the implement agent embeds artifacts in Linear comments
• Any time you have a local screenshot that needs to be in a Linear comment

The Iron Rule

Capture and upload are ONE atomic operation. NEVER capture a screenshot without uploading it to Linear in the same test item. If you defer uploads to "later," you will run out of turns and ALL evidence will be lost. This happened on ISSUE-123 (9 screenshots taken across 3 runs, 0 uploaded — all rejected).

The rhythm is: screenshot → resize → encode → upload → next item. Not: all screenshots → all uploads.

Why This Exists

Two failure modes that corrupt screenshot uploads:

1. Retina size: iOS Simulator screenshots are retina PNGs (1260x2730+, ~140KB+). Base64-encoding raw files produces ~190K chars that exceed context limits. Agents panic-resize to illegible thumbnails.

2. Cross-turn truncation: Even resized base64 (~18K chars) gets truncated by context compaction when the agent generates base64 in one API turn and calls create_attachment in the next. ISSUE-123 lost half its base64 this way (18,984 → 8,498 chars).

This skill uses a file-based pipeline — base64 is written to disk, never held in context across turns. The agent reads the file content in the same tool call that uploads it.

---

iOS Simulator Screenshots

Step 1: Capture

mcp__ios-simulator__screenshot(output_path: "verify-{item}.png")

Step 2: Resize + encode to file (SINGLE Bash call)

CRITICAL: Do this in ONE Bash command. Never let base64 sit in context across turns.

sips -Z 800 -s format jpeg verify-{item}.png --out verify-{item}.jpg 2>/dev/null && base64 -i verify-{item}.jpg > verify-{item}.b64 && wc -c < verify-{item}.b64