The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git commands (such as git worktree list, git worktree remove, and git show) to manage project workspaces. These operations are implemented using execFile with argument arrays, preventing shell injection vulnerabilities.
[DATA_EXPOSURE]: The skill reads AI agent transcript files from ~/.codex and ~/.claude. This access is essential to its primary purpose and the data is served only to a local-only server (127.0.0.1). The server includes an origin check to prevent unauthorized cross-origin requests from browsers.
[SAFE]: The UI is entirely self-contained; all fonts and frontend libraries (Preact, htm) are bundled or vendored locally to avoid external CDN dependencies and potential supply chain risks.
[SAFE]: Destructive actions, such as deleting sessions or removing worktrees, require explicit user confirmation through the browser UI and include validation checks to prevent accidental data loss (e.g., refusing to delete currently active sessions or dirty worktrees).

session-atlas