The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes instructions to download a shell script from a remote server and pipe it directly into a shell interpreter (curl -sSL http://agents.allium.so/cli/install.sh | sh). This is a critical security risk as it allows for arbitrary code execution on the host system from an unverified source.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch secondary markdown files containing further instructions from a remote domain (agents.allium.so) during execution. This dynamic loading of instructions from an external source bypasses static analysis and introduces risks of content tampering.
[COMMAND_EXECUTION]: The skill relies on the execution of a locally installed command-line interface tool (allium). Since this tool is installed via a remote script, its behavior is governed by the external script provider.
[PROMPT_INJECTION]: The skill exhibits a significant surface for indirect prompt injection by ingesting untrusted data from external URLs.
Ingestion points: Instruction sets are fetched from https://agents.allium.so/skills/x402-setup.md, x402-developer.md, and x402-explorer.md.
Boundary markers: There are no delimiters or explicit instructions provided to the agent to disregard or treat the fetched content as data rather than instructions.
Capability inventory: The skill possesses the ability to execute system commands via the allium CLI.
Sanitization: There is no evidence of validation, escaping, or filtering of the content retrieved from the remote URLs before it is processed by the agent.

allium-x402