The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the shell to execute gh (GitHub CLI), npm, and git commands for managing releases, performing builds, and handling deployments.
[EXTERNAL_DOWNLOADS]: Downloads and executes the claude-flow package from the NPM registry using npx. This is part of the vendor's intended functionality for swarm orchestration.
[REMOTE_CODE_EXECUTION]: Utilizes npx to run claude-flow and claude-flow@alpha which are vendor-specific tools used to perform the core release automation logic.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its data processing workflows.
Ingestion points: Data enters the system from external commit messages and Pull Request titles/labels via gh api and gh pr list commands.
Boundary markers: The skill does not define explicit delimiters (e.g., XML tags or triple quotes) to separate untrusted commit/PR data from agent instructions.
Capability inventory: Agents have access to high-privilege capabilities including Bash execution, file system modifications (Write), and GitHub repository management via API.
Sanitization: The provided documentation does not show any sanitization or validation logic to filter potentially malicious instructions embedded in commit history or PR descriptions.

github-release-management