The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git to extract commit history and file differences to provide context for the announcement, and it employs the gh CLI to interact with the GitHub GraphQL API for the publication phase.
[EXTERNAL_DOWNLOADS]: Fetches configuration and reference Markdown files from the author's own GitHub repository (https://raw.githubusercontent.com/levnikolaevich/claude-code-skills) using WebFetch as a fallback when local copies are unavailable.
[PROMPT_INJECTION]: The skill processes untrusted input from the repository (such as CHANGELOG.md and source code files) to compose its announcements, which presents an indirect injection surface. This is effectively mitigated by a mandatory user-approval gate and a multi-step fact-checking protocol. Ingestion points: Local repository files including CHANGELOG.md, README.md, and source code. Boundary markers: None explicitly defined in the prompt instructions. Capability inventory: Local file access, shell command execution, and network writing via the GitHub CLI. Sanitization: Relies on a verification phase (Phase 4) and a final human-in-the-loop review (Phase 5) before publication.

ln-912-community-announcer