show-env-status
Pass
Audited by Gen Agent Trust Hub on Feb 27, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill uses
mise envto display environment variables and attempts to redact secrets using a case-insensitive grep filter. - Evidence: The command
mise env 2>/dev/null | grep -v -i "TOKEN\\|KEY\\|SECRET\\|PASSWORD\\|CREDENTIAL"inSKILL.md. - Risk: This denylist approach is insufficient as it may miss sensitive variables using different naming conventions (e.g., AUTH, SSH_DATA, DATABASE_URL containing credentials, or custom application secrets).
- [COMMAND_EXECUTION]: The skill executes multiple local shell commands to gather information about the repository and tool configurations.
- Evidence: Usage of
mise,grep,ls, andechovia the Bash tool. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via untrusted repository metadata and file names.
- Ingestion points: File names via
lsand task definitions viamise tasks lsinSKILL.md. - Boundary markers: Absent; data is ingested without delimiters or specific instructions to the model to ignore embedded instructions.
- Capability inventory: Access to the
Bashtool enables command execution based on processed data. - Sanitization: Absent; outputs are not sanitized or validated before being presented to the model.
Audit Metadata