The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill possesses a significant attack surface by ingesting untrusted git commit messages and providing instructions to write the summarized output back to the file system (e.g., CHANGELOG.md).
Ingestion points: Git commit history and messages retrieved from the local repository via shell commands like git log.
Boundary markers: Absent. The instructions do not define delimiters or provide guidance to the agent to ignore or sanitize instructions found within the commit data.
Capability inventory: The skill description implies the agent has the capability to read git history and write files to the local directory.
Sanitization: Absent. The agent is encouraged to "translate" and "summarize" technical content, which involves deep processing of untrusted strings, increasing the likelihood of executing embedded prompt injection payloads.

changelog-generator