The Agent Skills Directory

[COMMAND_EXECUTION]: The asm-exec script facilitates command execution by acting as a wrapper that invokes subprocess.run on provided arguments. It also executes the local aws CLI to harvest credentials for request signing.
[CREDENTIALS_UNSAFE]: The utility retrieves AWS access keys and session tokens from the environment and local CLI configuration. This is a functional requirement for authenticating with the secret resolution backend and follows standard AWS credential provider patterns.
[EXTERNAL_DOWNLOADS]: The script communicates with an external AWS MCP endpoint (https://aws-mcp.us-east-1.api.aws/mcp) to resolve secret values. This endpoint is hosted on a well-known service domain and used for legitimate API interaction.
[DATA_EXFILTRATION]: While the tool's primary goal is to prevent data leakage, it creates a capability where resolved secrets are passed as arguments to other tools. If an agent is maliciously instructed to use network-active commands like curl with these references, secrets could be transmitted to external domains. However, the script implements single-pass substitution to prevent recursive expansion attacks.

aws-secrets-manager