The Agent Skills Directory

[DATA_EXPOSURE]: The initialization script scripts/better_auth_init.py explicitly targets and reads from .claude/.env and .claude/skills/.env. These paths are typically used by agent platforms to store internal configurations and credentials. Accessing these files exposes agent-level secrets to the skill's execution context.
[OBFUSCATION]: The skill distribution contains binary SQLite database files (scripts/.coverage and scripts/tests/.coverage). These are development artifacts that have no functional purpose for the skill and serve as non-textual data containers that could potentially hide payloads or leak developer environment metadata.
[COMMAND_EXECUTION]: The script scripts/better_auth_init.py performs interactive filesystem operations, including creating backups of existing .env files and writing new TypeScript configuration files (auth.ts). While intended for setup, these operations represent a significant capability for an unverified script.
[INDIRECT_PROMPT_INJECTION]: The scripts/better_auth_init.py script functions as an ingestion point for untrusted data via user input(). This data is interpolated into configuration files without robust sanitization or validation, creating a surface where malicious input could influence the generated authentication logic.
Ingestion points: User input via input() in scripts/better_auth_init.py and reading local .env files.
Boundary markers: Absent. The script writes directly to file templates.
Capability inventory: Filesystem writes (write_text), directory creation (mkdir), and environment reading (os.environ).
Sanitization: Minimal; only basic quote stripping is performed on environment values.

better-auth