The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes shell commands for git and gh (GitHub CLI). It employs a robust security model for this, including single-quote rendering of all dynamic values (branch names, titles) to prevent command injection and using --body-file with temporary files to avoid heredoc-based injection.- [DYNAMIC_EXECUTION]: The skill uses bash -p (privileged mode) for its sanitization script to ensure a clean execution environment, stripping caller-controlled environment variables and functions that could be used to bypass security checks. This is a defensive hardening measure.- [EXTERNAL_DOWNLOADS]: The skill mentions the github/gh-stack extension for the GitHub CLI. This is a well-known tool from a trusted organization (GitHub) used for its intended purpose of managing stacked PRs.- [DATA_EXFILTRATION]: No evidence of data exfiltration was found. Communication is limited to official GitHub APIs via the gh tool. The skill includes specific logic to avoid logging potentially sensitive content found in PR bodies during sanitization.- [PROMPT_INJECTION]: The skill uses instructional markers like "IMPORTANT" and "CRITICAL", but these are used for internal workflow logic and do not attempt to override the agent's core safety guidelines.

create-pr