The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of the 'tempo' CLI by downloading a script from https://tempo.xyz/install and immediately executing it using bash. This pattern allows for the execution of arbitrary remote code on the user's system without prior verification or integrity checks.
[PROMPT_INJECTION]: The skill contains instructions that explicitly forbid the agent from performing external validation, such as "Do NOT search for additional documentation" and "Do not search for other install instructions." These directives are designed to bypass the agent's ability to verify the safety or legitimacy of the installation process.
[COMMAND_EXECUTION]: The skill relies on the execution of multiple shell commands to manage a local wallet and interact with the tempo CLI, which increases the attack surface if the CLI itself or its inputs are compromised.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download files from URLs provided in API responses using curl -fsSL "<url>" -o <filename>. Because these URLs originate from external services, this behavior can be used to trick the agent into downloading and saving malicious payloads locally.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external API responses (tempo request) and service discovery metadata (tempo wallet -t services).
Ingestion points: API response bodies and service listing metadata from the tempo infrastructure.
Boundary markers: No specific boundary markers or "ignore" instructions are present to prevent the agent from obeying instructions that might be embedded in the external API responses.
Capability inventory: The skill has the capability to execute shell commands (tempo), write files to the local system (curl -o), and perform network requests.
Sanitization: No sanitization or validation logic is defined to inspect the content of external data before the agent acts upon it.

tempo