The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes the @useclawlink/cli tool via shell commands to perform authentication (login, connect) and interact with Humanloop resources (actions, describe, run).
[EXTERNAL_DOWNLOADS]: The setup process uses npx to download and execute the @useclawlink/cli package directly from the NPM registry. This is a standard deployment method for this vendor's tools.
[DATA_EXPOSURE]: The skill identifies that it stores authentication credentials locally at ~/.clawlink/credentials.json. This is standard behavior for CLI-based authentication but involves the creation of a sensitive file in the user's home directory.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data when reading experiments or sessions from Humanloop, which could theoretically contain malicious instructions.
Ingestion points: humanloop_list_experiments and humanloop_list_sessions tools described in SKILL.md.
Boundary markers: None present in the instructions to delimit external data.
Capability inventory: The skill can execute actions using npx @useclawlink/cli run based on agent decisions.
Sanitization: No explicit sanitization or validation of the data retrieved from the Humanloop API is mentioned.

humanloop