The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill facilitates the creation of Bash scripts. It includes specific safeguards, such as a 'fixture mode' for safe validation and instructions for the agent to verify script integrity using static analysis tools instead of direct execution.
[DATA_EXPOSURE_AND_EXFILTRATION]: The provided template contains functions to persist data to environment files and GitHub secrets. These functions are protected by environment-level opt-ins and mandatory interactive confirmation prompts, ensuring the human operator retains full control.
[INDIRECT_PROMPT_INJECTION]: The skill processes user input to define script stages. The potential for malicious instruction injection is mitigated by the 'disable-model-invocation' constraint and the necessity for human oversight and manual execution of the final authored script.

wizard