The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input which could potentially be used to manipulate underlying git commands or agent behavior.
Ingestion points: The $ARGUMENTS variable in SKILL.md accepts arbitrary user input intended for version or branch specifications.
Boundary markers: The workflow instructions do not utilize delimiters or specific instructions to the agent to ignore embedded commands within the $ARGUMENTS string.
Capability inventory: The skill is granted access to the Bash(git:*), Read, and Write tools, providing a surface for file and repository modification.
Sanitization: No validation, filtering, or escaping of the user-supplied arguments is performed in the skill definition.
Note: The skill references a workflow pipeline at ../../references/gitflow-finish-pipeline.md, which is outside the local skill directory.

finish-hotfix