The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a shell script to locate and execute a binary named 'wbso'. The search logic includes the current working directory ('$PWD'), which could result in the execution of an unintended or malicious file if the user runs the skill in an untrusted directory.
[COMMAND_EXECUTION]: User-provided feedback text is used as an argument in a shell command ('--message ""'). The instructions do not specify any escaping or sanitization for this input, creating a potential vector for command injection if the user input contains shell metacharacters like backticks or semicolons.
[DATA_EXFILTRATION]: The skill is designed to send user input to an external endpoint managed by WBSO.ai. While this is the intended purpose for feedback, it involves transmitting user-generated content to a remote service.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user feedback that is then used in a sensitive capability (shell execution).
Ingestion points: User feedback message collected in Step 2 of the instructions.
Boundary markers: None are used to encapsulate or neutralize the user input within the command string.
Capability inventory: Execution of the 'wbso' CLI tool via a shell environment in Step 3.
Sanitization: Absent; the instructions explicitly tell the agent to take the text 'literally' without modification.

feedback