The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the @catalyst-cloud/catalyst-skills package from the public npm registry using npx, npm, or bunx. This package originates from a source not recognized as a trusted organization in the current context.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary functionality, specifically using npx, npm install -g, and bunx to run the joining tool.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for command injection by taking untrusted user input and placing it directly into a shell command string.
Ingestion points: The skill instructions explicitly tell the agent to "Ask the human for their account key" (SKILL.md).
Boundary markers: No boundary markers or delimiters are defined for the input key.
Capability inventory: The skill utilizes shell execution capabilities (npx, npm, bunx) across its core operations.
Sanitization: There is no evidence of input validation or escaping for the <account-key> variable before it is interpolated into the shell command.
[CREDENTIALS_UNSAFE]: The skill handles an "account key" credential. While it provides instructions to secure the resulting configuration file (mode 0600), the practice of passing secrets as command-line arguments can lead to sensitive data exposure in process monitoring tools or shell history files.

join