The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute standard system and version control commands, including git log, git diff, and diff -ru, to manage local code changes. These commands operate on directories and project metadata provided by the user.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code and history from project directories, which presents an attack surface for indirect prompt injection. Ingestion points: Data is read from skills-lock.json and the .agents/skills/ directory. Boundary markers: No explicit delimiters are used to separate backported code from agent instructions. Capability inventory: The skill can read local files, generate diffs, and execute commits via the /git-commit tool. Sanitization: Content is not sanitized before it is summarized or committed back to the source repository.

backport-skill