The Agent Skills Directory

[COMMAND_EXECUTION]: The bash instruction in SKILL.md (Step 3) directs the agent to interpolate user-provided input into a shell command: "$WBSO_CLI" login --api-key "<KEY>". This creates a command injection surface.
Ingestion point: API key input provided by the user in Step 2.
Boundary markers: None present to delimit the input.
Capability inventory: Direct shell command execution.
Sanitization: No instructions are provided for the agent to sanitize or escape the input before interpolation.
[COMMAND_EXECUTION]: The SKILL.md file uses a find command to dynamically locate the wbso executable across various directories including $HOME/.codex/plugins/cache. Executing binaries found via wide path searches in shared cache directories could lead to the execution of unintended or malicious files if those directories are user-writable or shared between plugins.

auth