The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it frequently ingests and processes untrusted data from the Git repository environment while maintaining high-privilege capabilities.
Ingestion points: Git commit messages, branch names, and repository history accessed through git sl, git query, and git smartlog as detailed in SKILL.md and references/commands.md.
Boundary markers: Absent; the skill instructions do not require the use of delimiters or 'ignore' instructions when interpolating repository metadata into the agent's context.
Capability inventory: Extensive shell execution via git and git-branchless CLI tools, including the highly capable git test run --exec command which executes arbitrary shell strings across commit stacks.
Sanitization: Absent; there are no defined validation or escaping procedures for strings retrieved from the repository before they are processed by the agent.
[COMMAND_EXECUTION]: The core functionality of the skill involves the execution of various Git-related shell commands. While appropriate for the skill's stated purpose of version control management, it requires the agent to have broad execution permissions within the user's shell environment.

git-branchless