Skill

agent-capture

Record, screenshot, and convert macOS windows and screens for AI agent evidence

Verified: 2026-05-13 (printing-press-ingest-2026-05-13+enrich-capability-skill)

When to use agent-capture

Choose if

You're on macOS 12.3+ and want an agent to produce visual evidence — screenshots, window recordings, GIFs, before/after diffs, simulated demos — without piloting a UI automation framework. Pairs well with PR-writing agents that need to attach proof of a UI change.

Avoid if

You're not on macOS, or you can't grant Screen Recording permission to the terminal running the agent. The CLI also has no remote-host story — it captures whatever screen the local macOS session owns, so headless CI runners and Linux dev containers are out of scope.

Risk Flags

  • MEDIUM runtime macOS-only (12.3+). No Windows or Linux build path — agents on other hosts get nothing.
  • MEDIUM auth First-time Screen Recording permission requires a terminal restart before captures work; agents that try to record before the restart get a permission-denied error.
  • LOW runtime Optional deps (ffmpeg, VHS, Remotion) are not bundled. Video and GIF conversion fail when ffmpeg is missing; demo rendering is skipped without VHS/Remotion.
  • LOW scope GIF conversion may auto-skip frames and shrink dimensions to honor --max-size; the output is not guaranteed pixel-faithful to the original recording.

Cost

Type: Free

Install

Default

go install github.com/mvanhorn/printing-press-library/library/developer-tools/agent-capture/cmd/agent-capture-pp-cli@latest
# First time only: grant Screen Recording permission to your terminal,
# then restart the terminal so the permission takes effect.

Setup docs →

Estimated time to first success: ~5 min

Dependencies

Minimum runtime: macOS 12.3+ with Xcode Command Line Tools; ffmpeg required for video/GIF conversion; VHS and Remotion optional for demo rendering

Distribution

Repository
https://github.com/mvanhorn/printing-press-library
License
MIT