The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes dynamic context injection placeholders (!) to execute shell commands when the skill is loaded. These commands incorporate user-supplied arguments (such as $1 for the base branch) directly into shell strings. This introduces a potential command injection surface where a crafted branch name containing shell metacharacters (e.g., backticks or semicolons) could lead to arbitrary command execution in the local environment.
[PROMPT_INJECTION]: The skill processes untrusted data from the local repository (via git status, git log, and git diff) to generate an organization plan. This creates a surface for indirect prompt injection where malicious instructions hidden in source code or commit messages could attempt to override the agent's logic.
Ingestion points: Repository metadata and diffs are gathered using git status, git log, and git diff within the SKILL.md dynamic context.
Boundary markers: None; external content is processed directly to form the categorization plan.
Capability inventory: The skill can perform file system operations via git reset, git add, and git commit commands.
Sanitization: No specific sanitization or filtering is applied to the repository output before it is analyzed by the agent.

organize-commits