The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the execution of shell commands such as git (for branch, tag, and log management), gh (for GitHub Pull Request and API operations), and bun (for local automation scripts). These commands are necessary for the skill's primary function and are used with specific, restricted arguments.
[PROMPT_INJECTION]: The skill processes untrusted external data from git commit subjects and GitHub PR metadata, creating a surface for indirect prompt injection.
Ingestion points: Commit history is ingested via git log and PR details via gh pr view as defined in references/release-notes-style.md.
Boundary markers: Explicit boundary markers for external content are not utilized, though the generated output follows a strict markdown template.
Capability inventory: The agent can perform shell command execution and file operations.
Sanitization: The skill significantly reduces risk by using strict regex patterns (grep -oE '$#[0-9]+$$') to extract only numeric PR IDs and employing jq for parsing JSON responses from the GitHub CLI, ensuring that data is handled as structured text rather than executable commands.

version-release