The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to orchestrate internal utility scripts for distilling sessions, scanning for secrets, and generating the standalone skill package. It also executes npx playwright to launch browser sessions for recording and replaying UI steps.
[EXTERNAL_DOWNLOADS]: The recording and UI replay scripts (scripts/record_session.sh and scripts/replay_ui.sh) install the playwright package from the official NPM registry and may download required browser binaries or ffmpeg. These are well-known tools provided by Microsoft and the open-source community for automation and media processing.
[DATA_EXPOSURE]: While the skill's primary function involves handling authenticated web sessions (cookies and tokens), it includes several security controls: a secret scanner (scripts/scan_secrets.py) that identifies sensitive data like payment cards or private keys, and a dedicated 'setup' mode (--with-setup) that strips credentials from generated skills to ensure they can be shared without exposing the original user's session.
[SAFE]: The skill implements proactive security measures such as redacting credential values in distilled lessons, using isolated Chrome profiles to prevent session pollution, and requiring explicit user confirmation before executing any state-changing (mutating) web requests during replay.

generate-web-skills