The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches commit history and project metadata from well-known and official sources, including chromiumdash.appspot.com, googlesource.com, cr-rev.appspot.com, and api.github.com.
[COMMAND_EXECUTION]: Employs a local Node.js script (scripts/changelog-gen.js) to programmatically query repository APIs using the system shell.
[PROMPT_INJECTION]: The skill processes untrusted commit messages from external repositories, creating an indirect prompt injection surface. Ingestion point: Commit messages fetched via scripts/changelog-gen.js. Boundary markers: Data is structured as JSON, but lacks explicit injection-prevention delimiters. Capability inventory: Agent capabilities are limited to summarization and formatting; no sensitive file access or network write operations are enabled. Sanitization: No content filtering is applied to the retrieved commit messages.

changelog-generation