send-notification
Warn
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to run TypeScript files (
bun pushover_core.ts) and helper scripts (_lib/resolve_pushover_secret.sh) to perform its primary function. - [DATA_EXFILTRATION]: Instructions explicitly use
env -u HTTPS_PROXY -u HTTP_PROXYto bypass the execution environment's proxy settings. This technique is used to ensure direct communication with the Pushover API but can be used to evade network-level security controls or monitoring. - [CREDENTIALS_UNSAFE]: The skill relies on an external script (
resolve_pushover_secret.sh) to retrieve API tokens from 1Password or the macOS Keychain. While this avoids hardcoding secrets, it requires the agent to have access to these credential management systems.
Audit Metadata