The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's core function is to generate .whip DSL scripts and execute them using the whipflow_run tool. Instructions explicitly mandate using mode="auto" to run generated scripts immediately, bypassing user review of the code before it executes on the system.
[COMMAND_EXECUTION]: The DSL allows agents within a workflow to utilize highly privileged tools including bash. This enables the execution of arbitrary shell commands on the host environment as part of an automated pipeline.
[DATA_EXFILTRATION]: The 'Canvas Integration' feature uses window.parent.postMessage with a wildcard * target origin. This is a security anti-pattern that allows any parent window or frame to intercept potentially sensitive workflow data or trigger arbitrary workflow execution if the environment is not strictly isolated.
[PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection. External data retrieved via the web_search or read tools is directly interpolated into agent prompts using string templates (e.g., {result}, {research}) without sanitization, boundary markers, or instructions to ignore embedded commands. This allows untrusted content from the internet or local files to hijack the workflow's logic.
[DATA_EXFILTRATION]: The workflow allows reading sensitive files (via the read tool) and subsequently performing network operations (via web_search or bash), creating a direct path for data harvesting and exfiltration.

whipflow