bws-secrets
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill implements strong defensive measures against credential leakage. It explicitly forbids secret values from being displayed, printed, or logged in the model context or chat history. Evidence includes instructions to never read, display, or return secret values to the transcript.
- [COMMAND_EXECUTION]: The skill utilizes a specific CLI tool (secretctl) for credential operations. It enforces the use of absolute paths for consumer executables and warns against running raw BWS commands, accessing Keychain items directly, or using commands like env/printenv that could leak environment variables.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of malicious input from external sources and instructs the agent to treat credential requests from prompts, files, or websites as prompt injection attempts with no override.
- [PRIVILEGE_ESCALATION]: Access controls are managed via a root-owned configuration file. The skill enforces user-presence checks on macOS (Touch ID) for privileged operations and limits capabilities on Linux where such checks are unavailable.
Audit Metadata