The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data through git diff and git log. Instructions embedded within code comments or commit messages in the repository could potentially influence the agent's behavior during the commit process.
Ingestion points: The skill reads git diff (file content changes) and git log (previous commit history).
Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard instructions found within the diff or logs.
Capability inventory: The agent has the capability to execute git add, git commit, and git push.
Sanitization: There is no evidence of filtering or sanitizing the content retrieved from git commands before it is processed by the agent.
[METADATA_POISONING]: There is a direct contradiction between the skill's description and its Process section. The description states the skill is for creating commits "without pushing," whereas Step 8 of the process explicitly instructs the agent to "Run git push unless otherwise specified." This discrepancy could lead the agent to perform an action (uploading code) that contradicts the user's expectations based on the skill's title and description.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary functions. It uses git to stage files, create commits, and push changes to remote repositories. While these are intended functions, they represent the capability to modify the project state and interact with remote servers.

commit