The package exposes an explicitly invoked skill installer and user-run tools. The flagged environment values become configured browser headers, while usage reporting reads local usage metadata; the reviewed source showed no unsolicited exfiltration or harmful automatic action.
Static reason
One or more suspicious static signals were detected.
Trigger
The installer and usage scripts run only when explicitly invoked; token guards require the --guards option.
Impact
The observed behavior is scoped to the selected project or user-requested browser automation; no unrelated data collection or credential transmission was identified.
Mechanism
The installer places skills and workflow guidance in the selected project. Optional guards add a local hook. Playwright helpers use configured browser headers and can probe localhost when called.
Evidence
selected project native skill directoryselected project workflow guidance fileselected project/.claude/hooks/papi-token-guards.mjs (with --guards)selected project/.claude/settings.json (with --guards)skills/playwright-skill/.temp-execution-<timestamp>.js (during runner use)