The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes functionality to read sensitive local chat history from directories such as ~/.claude/projects/ and ~/.codex/sessions/ using scripts/harvest.py. While used for content extraction, these transcripts are highly sensitive and can contain private code, tokens, or credentials.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx --yes hyperframes@0.8.71 in its workflow, which automatically downloads and executes code from the npm registry. Additionally, scripts/setup.sh performs an npm ci install from a provided lockfile.
[DYNAMIC_EXECUTION]: The script scripts/schedule.mjs uses the node:vm module to dynamically load and execute the logic in assets/engine/scenes/compile.js within a sandboxed context.
[COMMAND_EXECUTION]: The skill frequently executes shell commands to coordinate the animation and rendering process, including calls to ffmpeg, python3, node, npx, and swift (on macOS).

session-story