The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on executing shell commands via the Bash tool to perform Git operations (git worktree, git branch, git rebase, git push) and GitHub CLI interactions (gh pr list, gh repo view). This is the intended primary purpose of the skill.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of the gh-stack extension from GitHub's official repository (github/gh-stack) if it is missing. This download targets a trusted organization and is a standard part of the gh toolchain environment setup.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted data from external sources, such as branch names, issue titles, and pull request metadata fetched via git and gh tools.
Ingestion points: Data enters the agent context through git branch -vv, git worktree list, and gh pr list commands (referenced in clean.md and SKILL.md).
Boundary markers: The skill lacks explicit boundary markers for external metadata, though it uses kebab-casing and slugification for branch names which acts as a form of structural sanitization.
Capability inventory: The agent has the capability to execute shell commands (Bash tool) and write to the local filesystem (during conflict resolution).
Sanitization: Branch names are sanitized by flattening slashes to dashes and enforcing kebab-case slugs, which prevents basic path traversal or command injection via malformed branch names.
[DATA_EXFILTRATION]: The skill performs network operations using git push to origin and gh API calls to GitHub. These operations are limited to standard repository synchronization and pull request management as requested by the user.

gitkit