What problem does it solve? Marketing and README screenshots of a desktop app go stale after every UI change, and retaking them against real user data risks leaking private sessions, tokens, and file paths. This Skill regenerates the full docs/screenshots set from a fully synthetic, anonymised dataset without ever touching the real ~/.agento or ~/.claude directories. ## Core Features & Use Cases - Synthetic corpus generation: gen_corpus.py deterministically writes ~300 Claude Code session JSONL files across 10 invented projects, including sub-agent transcripts, token usage, and a scripted hero session for detail shots. - Isolated app launch: launch.sh runs the Tauri dev build against a fake HOME (~/.agento-demo) under XWayland so the window can be sized to the README frame, while seed_api.py and seed_db.py populate agents, integrations, scheduled tasks, chats, and job history. - Deterministic screenshot capture: shoot.sh drives the real webview through the ui-verify helper, waiting for settle predicates before each capture so React re-renders never race the snapshot. - Use Case: After a UI redesign makes docs/screenshots stale, run the five-step pipeline (corpus, launch, seed, shoot, clean) to regenerate every light-theme PNG in docs/screenshots/light with consistent, anonymised data. ## Quick Start Ask the agent to regenerate the README screenshots with demo data, and it will build the synthetic corpus, launch the isolated app, seed it, and shoot the full screenshot set into docs/screenshots/light.
