The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a localized Python script, scripts/changelog.py, to manage the state of the changelog and render the final HTML bundles.
[COMMAND_EXECUTION]: It calls shell scripts from a sibling asset-manager skill to perform secure uploads and downloads of website artifacts to the platform.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text or JSON data to generate public entries. It includes comprehensive security logic to mitigate injection risks:
Ingestion points: User-provided changelog entries are written to a temporary new.json file.
Boundary markers: Data is processed through structured JSON schemas rather than raw string interpolation.
Capability inventory: The skill has the ability to read/write files and perform network operations via separate asset-management scripts.
Sanitization: The changelog.py script employs an escape() function to prevent HTML injection and uses multiple regular expressions to identify and block volatile content, including <script>, <iframe>, link, and inline event handlers, as well as external resource fetches.

public-changelog