The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute git and gh (GitHub CLI) commands for inspecting repository state, managing branches, and creating pull requests. These operations are restricted to the local environment and the configured repository origin.
[PROMPT_INJECTION]: The skill is assessed for indirect prompt injection surfaces as it ingests and processes untrusted repository data.
Ingestion points: Reads conventions from local and repository-level configuration files (CLAUDE.md, AGENTS.md) and analyzes the content of git diff outputs in SKILL.md.
Boundary markers: Implements quoted HEREDOCs (cat <<'EOF') when piping content into shell commands for commit messages and PR bodies, which prevents the shell from interpreting malicious content as commands (shell injection).
Capability inventory: The skill uses Bash (command execution), Read (file access), Grep, and Glob tools.
Sanitization: Features a mandatory blocking gate in Step 2 that requires explicit user confirmation before proceeding if discrepancies are found, ensuring that the AI agent does not automatically act on potentially manipulated data without oversight.

ship-pr