The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes dynamic context injection (the ! command syntax) to execute shell commands during the skill loading phase. It interpolates user-provided arguments (referenced as ${1}) directly into the git log and git diff shell commands. This pattern is vulnerable to command injection, as shell metacharacters in the input could allow for the execution of arbitrary commands on the host system.
[PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data sources, specifically commit messages and diff statistics. These sources are attacker-controllable and could contain malicious instructions designed to manipulate the agent's behavior during the PR body generation phase (Indirect Prompt Injection).
[COMMAND_EXECUTION]: The instructions explicitly direct the agent to perform actions like git push and gh pr create without seeking user confirmation. This removal of human oversight amplifies the impact of the command and prompt injection vulnerabilities by allowing potentially manipulated or malicious PRs to be created and pushed automatically.

pr