The Agent Skills Directory

[PROMPT_INJECTION]: The workflow instructions (Step 5 and Step 6) explicitly direct the agent to execute terminal commands 'automatically' and state that 'no confirmation [is] needed'. This is an attempt to override standard safety guardrails that ensure human oversight for shell command execution.
[COMMAND_EXECUTION]: The skill instructs the agent to perform a series of terminal operations, including git status, git diff, and git commit. When combined with the instruction to skip user confirmation, this creates a pattern where the agent could autonomously execute commands based on potentially untrusted input (such as the content of the changes being committed).

conventional-commit