The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The documentation provides a command to fetch a shell script from a remote URL (https://raw.githubusercontent.com/zep-us/zeude/main/install.sh) and pipe it directly into bash. This pattern bypasses manual review of the script's contents and executes code from a repository that is not part of the established vendor ecosystem for this skill.
[COMMAND_EXECUTION]: The skill utilizes a 'Zeude Shim' designed to intercept and wrap the primary CLI commands. This mechanism could potentially be used to monitor user input or modify command behavior without explicit oversight.
[INDIRECT_PROMPT_INJECTION]: The skill implements hooks (UserPromptSubmit) that analyze user prompts at runtime to deliver context-aware suggestions. This creates an ingestion surface where untrusted data (the prompt) is processed to determine agent actions (skill deployment).
Ingestion points: User prompts via hook analysis.
Boundary markers: None specified in the instructions.
Capability inventory: Access to Bash, Write, Read, and the ability to sync configurations to ~/.claude/.
Sanitization: No evidence of sanitization or intent-validation for the processed prompts.

zeude