The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill processes untrusted code content from the repository and has the capability to modify the repository state through commits.
Ingestion points: Untrusted data enters the context via git diff and git status (SKILL.md).
Boundary markers: No specific delimiters are used to wrap the diff content to prevent it from being interpreted as instructions by the agent.
Capability inventory: The skill can execute git add and git commit via the Bash tool (SKILL.md).
Sanitization: There is no evidence of sanitization or validation of the diff output before it is processed for message generation.
[Command Execution] (LOW): The skill utilizes the Bash tool to execute git commands.
Evidence: Use of git diff, git status, git add, and git commit (SKILL.md).
Mitigation: The skill includes a 'Git Safety Protocol' which explicitly prohibits destructive commands and the updating of git configurations, reducing the risk of accidental harm.

git-commit