The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess.run method in scripts/draft_entries.py and various test scripts to execute git and python3. These executions are performed securely by passing arguments as a list and not using a shell, which prevents command injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted commit messages and PR titles from the repository to generate changelog drafts. This represents a vulnerability surface where maliciously crafted Git data could attempt to influence the agent's behavior during the review process.
Ingestion points: Untrusted data enters via git log output processed in scripts/draft_entries.py.
Boundary markers: None; commit messages are interpolated directly into Markdown lists in the drafted output.
Capability inventory: The skill has the capability to execute git commands and write to the local filesystem (e.g., updating CHANGELOG.md).
Sanitization: None; the Python logic extracts segments from commit messages using regex but does not perform escaping or filtering of the content.

update-changelogs