The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git and gh (GitHub CLI) commands to perform repository operations such as staging, committing, pushing, and inspecting history. Subprocess calls in the accompanying Python scripts are implemented using argument lists rather than shell strings, preventing shell-level injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes commit messages that may incorporate untrusted data. It implements strong 'Heredoc Safety Rules' for Bash/Zsh and 'PowerShell Here-String Rules' to ensure messages are treated as literal data, preventing command injection during the commit process.
Ingestion points: Commit messages drafted from user input and repository metadata.
Boundary markers: Uses quoted heredocs ('EOF') and single-quoted here-strings to wrap commit messages.
Capability inventory: File system access via git, network access via gh api, and process execution via subprocess.run.
Sanitization: Input to the wrap_message.py script is processed via regex to ensure only expected identifiers (e.g., repository names, issue numbers) are used in API calls.
[EXTERNAL_DOWNLOADS]: The wrap_message.py script performs network lookups to the GitHub API via the gh CLI to resolve issue and pull request references. This targets well-known official GitHub infrastructure.

commit