pp-agent-capture — macOS skill: window screenshots, recording
pp-agent-capture captures windows, records video, and converts to GIF from the terminal on macOS. Uses ScreenCaptureKit for window-level targeting, bundles evidence for PRs, and monitors UI changes.
Apache-2.0 306 installs ★ 2,142 updated 2026-10-24
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a claude-codeWhat pp-agent-capture does
This skill drives the agent-capture-pp-cli binary to capture specific windows or apps, record video, convert video to optimized GIF, diff against baseline screenshots, and bundle full evidence packages for pull requests. Under the hood: ScreenCaptureKit for precise window-level capture (what ffmpeg cannot do), ffmpeg for conversion, macOS Vision for OCR text extraction from windows.
Requires macOS and Screen Recording permission, which the skill requests on first run. The binary installs to $HOME/.local/bin (macOS/Linux) or %LOCALAPPDATA%\Programs\PrintingPress\bin (Windows) and must be on $PATH. If npx install fails, fall back to direct Go install (requires 1.26.6 or newer).
Includes screenshot and batch for single or multiple windows, record for video capture, pipeline for full capture-convert cycle in one go, diff for baseline comparison, evidence for PR bundles, find for fuzzy window matching, ocr for text extraction, and watch for periodic monitoring. All commands support --agent for structured JSON output.
What it helps with
- capture a screenshot of a specific window or macOS app
- record video of a screen region with minimal file size
- convert video to an optimized GIF with two-pass palette
- bundle before-and-after screenshots, video, and GIF for a PR
- find the right window by fuzzy-matching its title
Example prompts for the agent
- Take a screenshot of the Safari window
- Record the last 30 seconds of screen activity and convert it to GIF
- Create an evidence bundle with before-and-after screenshots and a video recording for the PR
How to install pp-agent-capture
1Skills CLI (Vercel)
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a claude-code2Manually: put the files in place
~/.claude/skills/pp-agent-capture/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/pp-agent-capture/
1Skills CLI (Vercel)
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a codex2Manually: put the files in place
.agents/skills/pp-agent-capture/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a cursor2Manually: put the files in place
.cursor/skills/pp-agent-capture/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install mvanhorn/printing-press-library cli-skills/pp-agent-capture/SKILL.md2Manually: put the files in place
.github/skills/pp-agent-capture/SKILL.md1Skills CLI (Vercel)
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a gemini-cli2Platform CLI
gemini skills install https://github.com/mvanhorn/printing-press-library --path cli-skills/pp-agent-captureManually: put the files in place
.devin/skills/pp-agent-capture/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about pp-agent-capture
How do I install pp-agent-capture in Claude Code?
Run npx skills add mvanhorn/printing-press-library --skill pp-agent-capture -a claude-code in the terminal, then verify the binary with agent-capture-pp-cli --version. The binary must be on $PATH.
What does pp-agent-capture do?
It captures windows, records video, converts to GIF, diffs screenshots against baselines, and bundles evidence for pull requests. ScreenCaptureKit enables window-level capture precision that ffmpeg cannot achieve.
Which platforms does pp-agent-capture support?
macOS only. The skill relies on ScreenCaptureKit, so it will not work on other operating systems.
Source text
SKILL.md · Apache-2.0
---
name: pp-agent-capture
description: "macOS screen capture, window recording, GIF conversion, and agent evidence bundles from the terminal. Built on ScreenCaptureKit for window-level targeting ffmpeg cannot do. Use when the user wants a screenshot of a specific window or app, a screen recording, a GIF conversion, a before/after diff, an evidence bundle for a PR, OCR text from a window, a terminal VHS recording, a Remotion render, or wants to watch a UI for changes. Requires macOS Screen Recording permission on first run."
author: "Matt Van Horn"
license: "Apache-2.0"
argument-hint: "<command> [args] | install cli"
allowed-tools: "Read Bash"
metadata:
openclaw:
requires:
bins:
- agent-capture-pp-cli
install:
- kind: go
bins: [agent-capture-pp-cli]
module: github.com/mvanhorn/printing-press-library/library/developer-tools/agent-capture/cmd/agent-capture-pp-cli
---
<!-- GENERATED FILE — DO NOT EDIT.
This file is a verbatim mirror of library/developer-tools/agent-capture/SKILL.md,
regenerated post-merge by tools/generate-skills/. Hand-edits here are
silently overwritten on the next regen. Edit the library/ source instead.
See the repository agent guide, section "Generated artifacts: registry.json, cli-skills/". -->
# Agent Capture - Printing Press CLI
## Prerequisites: Install the CLI
This skill drives the `agent-capture-pp-cli` binary. **You must verify the CLI is installed before invoking any command from this skill.** If it is missing, install it first:
1. Install via the Printing Press installer. It defaults binaries to `$HOME/.local/bin` on macOS/Linux and `%LOCALAPPDATA%\Programs\PrintingPress\bin` on Windows:
```bash
npx -y @mvanhorn/printing-press-library install agent-capture --cli-only
```
2. Verify: `agent-capture-pp-cli --version`
3. Ensure the reported install directory is on `$PATH` for the agent/runtime that will invoke this skill.
If the `npx` install fails (no Node, offline, etc.), fall back to a direct Go install (requires Go 1.26.6 or newer):
```bash
go install github.com/mvanhorn/printing-press-library/library/developer-tools/agent-capture/cmd/agent-capture-pp-cli@latest
```
If `--version` reports "command not found" after install, the runtime cannot see the binary directory on `$PATH`. Do not proceed with skill commands until verification succeeds.
## When to Use This CLI
Reach for this when the user wants:
- screenshot a specific window or app (`screenshot`, `batch` for multiple)
- record video of a window, app, display, or region (`record`)
- convert a recording to an optimized GIF (`convert`)
- do a full capture + record + GIF pipeline in one command (`pipeline`)
- diff against a baseline screenshot (`diff`) for before/after evidence
- bundle screenshots + recording + GIF as evidence for a PR or bug report (`evidence`)
- find the right window by fuzzy-matching its title (`find`)
- stitch multiple screenshots into an animated GIF (`stitch`)
- extract text from a window using macOS Vision OCR (`ocr`)
- record a terminal session via VHS tape files (`vhs`)
- render Remotion compositions to video or stills (`remotion`)
- monitor a UI by periodic screenshots (`watch`)
- save and replay capture configs (`preset`)
Skip it on non-macOS hosts; the CLI uses ScreenCaptureKit (macOS only). On first run it will prompt for Screen Recording permission; the `permissions` command guides that flow.
## Argument Parsing
Parse `$ARGUMENTS`:
1. **Empty, `help`, or `--help`** -> show `agent-capture --help`
2. **Starts with `install`** -> CLI installation (no MCP server ships today)
3. **Anything else** -> Direct Use (map to the best command and run it)
## Direct Use
1. Check installed: `which agent-capture`. If missing, offer CLI installation.
2. If permissions aren't granted, run `agent-capture permissions` first.
3. Use `agent-capture list` to see available capture targets (open windows, displays).
4. Use `agent-capture find <text>` to fuzzy-match a window title before capturing.
5. Execute with `--json` for structured output (agent-native default):
```bash
agent-capture <command> [args] --json
```
## Notable Commands
| Command | What it does |
|---------|--------------|
| `screenshot` | Capture a window, app, display, or region |
| `record` | Record video of a window, app, display, or region |
| `pipeline` | Record + convert + optimize in one command |
| `convert` | Video -> optimized GIF (two-pass palette) |
| `diff` | Capture + diff against a baseline |
| `evidence` | Full bundle (screenshots + recording + GIF) for a PR |
| `batch` | Screenshot multiple apps in one invocation |
| `find` | Fuzzy search open window titles |
| `list` | List available capture targets |
| `ocr` | Extract text from a window using macOS Vision |
| `stitch` | Combine screenshots into an animated GIF |
| `vhs` | Run a VHS tape file for terminal recording |
| `remotion` | Render Remotion compositions |
| `watch` | Periodic capture for UI monitoring |
| `preset` | Save / load capture configs |
| `permissions` | Guide Screen Recording permission setup |
| `health` | Machine-readable CI / agent preflight |
Run any command with `--help` for full flag documentation.
## Agent Mode
Add `--agent` to any command. Expands to: `--json --compact --no-input --no-color --yes`.
- **Pipeable** — JSON on stdout, errors on stderr
- **Filterable** — `--select` keeps a subset of fields, with dotted-path support (see below)
- **Previewable** — `--dry-run` shows the request without sending
- **Cacheable** — GET responses cached for 5 minutes, bypass with `--no-cache`
- **Non-interactive** — never prompts, every input is a flag
### Filtering output
`--select` accepts dotted paths to descend into nested responses; arrays traverse element-wise:
```bash
<cli>-pp-cli <command> --agent --select id,name
<cli>-pp-cli <command> --agent --select items.id,items.owner.name
```
Use this to narrow huge payloads to the fields you actually need — critical for deeply nested API responses.
## Exit Codes
| Code | Meaning |
|------|---------|
| 0 | Success |
| 2 | Usage error (wrong arguments) |
| 3 | Target not found (no matching window or display) |
| 4 | Permissions missing (Screen Recording not granted) |
| 5 | Capture error (ScreenCaptureKit failure, ffmpeg failure) |
© Matt Van Horn · Apache-2.0 · mvanhorn/printing-press-library