The Agent Skills Directory

[SAFE]: Technical analysis of the skill's scripts, reference documents, and configuration reveals no malicious patterns or security risks. The skill is functionally transparent and adheres to its stated purpose of managing GitHub resources.
[COMMAND_EXECUTION]: The skill provides several bash scripts in the scripts/ directory for automating complex GitHub tasks. These are well-documented and safe. Documentation in references/workflow-authoring.md includes commands like sudo rm -rf /usr/share/dotnet, which are correctly described as standard practices for freeing disk space on remote GitHub-hosted runners and are not executed locally.
[EXTERNAL_DOWNLOADS]: The skill references the skilz universal installer, which is authored by the same vendor (SpillwaveSolutions). This is considered a legitimate vendor resource for skill lifecycle management across different agent platforms.
[PROMPT_INJECTION]: The skill interacts with external, potentially untrusted data from GitHub (e.g., PR descriptions, issue bodies, workflow logs), representing a surface for indirect prompt injection.
Ingestion points: Data enters the context via gh search, gh pr list/view, gh issue list/view, and gh run view --log commands.
Boundary markers: No explicit markers or specific instructions to ignore embedded commands in external data are provided.
Capability inventory: The skill is authorized to use Bash for script execution and has full access to the gh tool for repository and automation management.
Sanitization: The skill uses jq to parse structured JSON data, which provides basic data handling, but content within the fields is not sanitized before presentation to the agent.

mastering-github-cli