The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from the local repository.
Ingestion points: Reads repository state and file changes using git diff and git diff --cached as described in SKILL.md.
Boundary markers: Absent; file content is analyzed directly to generate slugs and descriptions.
Capability inventory: Can execute git commit and git push via the Bash tool.
Sanitization: Employs HEREDOC formatting for commit message construction to prevent shell injection, though it does not filter the logical content for prompt injection patterns.
[DATA_EXFILTRATION]: The skill transmits data over the network via git push to a remote repository.
The operation targets the origin remote configured in the user's environment.
This functionality is essential to the skill's primary purpose.
Risk is mitigated by requiring user confirmation through AskUserQuestion before any push is executed.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform repository management tasks, including branch manipulation, stashing, and rebasing. It includes safety logic to exclude specific patterns of sensitive files (e.g., .env, *.pem, *.key) from being staged or committed.

commit