The Agent Skills Directory

[COMMAND_EXECUTION]: The installation instructions for the Twilio CLI utilize sudo to move binaries into /usr/local/bin. This pattern requests administrative privileges, which can lead to privilege escalation if executed by an agent in a sensitive environment.
[EXTERNAL_DOWNLOADS]: The skill instructs the download of the Twilio CLI binary from an Amazon S3 bucket (twilio-cli-prod.s3.amazonaws.com). While Twilio is a well-known service provider, downloading and installing binaries from remote sources at runtime presents a supply chain risk.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection (Category 8) where untrusted phone number data is interpolated into API requests and shell commands.
Ingestion points: The phone variable is used across Python, Node.js, and Bash examples (e.g., client.lookups.v2.phoneNumbers(input).fetch()).
Boundary markers: The examples lack boundary markers or delimiters to isolate user input from the rest of the command or URL.
Capability inventory: The skill utilizes network operations (curl, SDKs) and CLI tools that could be manipulated through crafted inputs.
Sanitization: Code snippets lack explicit input validation or sanitization for the phone variable, potentially allowing parameter pollution or command manipulation.
[DATA_EXFILTRATION]: The skill references sensitive file paths for configuration and secrets, such as /run/secrets/twilio_api_key_secret and /etc/openclaw/twilio/lookup.toml. Interaction with these paths involves the exposure of sensitive system areas and credential storage.

twilio-lookup