The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing shell commands via the gh (GitHub CLI) and git binaries to manage branches.
The $issue variable is interpolated directly into gh issue view "$issue" in workflows/issue-branch.md, which could lead to command injection if the user provides a malicious identifier string.
The skill executes git checkout, git fetch, and git rebase using variables derived from external issue data.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it processes data retrieved from external GitHub issues.
Ingestion points: Data enters the agent's context from the gh issue view command output, specifically the issue title and number (found in workflows/issue-branch.md).
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are used when processing the issue metadata.
Capability inventory: The skill has the capability to execute shell commands (git checkout, git fetch, git rebase) and manipulate the local file system's git state.
Sanitization: The workflow includes a robust sanitization step in workflows/issue-branch.md (Step 2) that transforms the issue title into a kebab-case branch name (lowercase, alphanumeric only, hyphen-separated). This effectively prevents shell injection through the issue title in subsequent commands.

new-branch