The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted release notes through the BODY input, creating a potential surface for indirect prompt injection.
Ingestion points: The BODY input field in SKILL.md receives raw markdown data.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present.
Capability inventory: The agent is authorized to write to the docs/changelogs/ directory and execute npm run format.
Sanitization: The skill implements specific formatting rules, such as transforming PR URLs and removing a 'New Contributors' section, which provides some structure to the processed data.
[COMMAND_EXECUTION]: The skill utilizes npm run format as a finalization step. This is a common development task for maintaining file formatting and is considered a safe, routine operation.
[EXTERNAL_DOWNLOADS]: The skill's templates include documentation for installing the Gemini CLI via npm install and provide links to the project's official repository on GitHub. These references point to well-known and reputable project resources.

docs-changelog