The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script scripts/rewrite_commit_messages.py to interact with Git. The script uses subprocess.run with list-based arguments and explicitly avoids shell=True, which protects against shell injection. The script includes robust safety logic: it requires a clean repository state, verifies that the target branch is not protected (e.g., main or release branches), and validates that the final state of the code matches the original head exactly.
[DATA_EXFILTRATION]: No network operations or external data transfer mechanisms were found. The skill operates entirely on the local Git repository and does not access sensitive system directories, environment variables, or remote servers.
[PROMPT_INJECTION]: The instruction set in SKILL.md establishes clear 'Non-negotiables' and safety-first policies. It explicitly prohibits the agent from inventing details, overriding safety filters, or performing dangerous operations like force-pushing, instead providing deterministic checks in the helper script.

commit-history-authoring