The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Git commands and a local bash script (scripts/inspect-operation.sh) to perform its primary function of history integration and state inspection.
[COMMAND_EXECUTION]: The helper script scripts/inspect-operation.sh implements defensive shell programming, including the use of set -euo pipefail for error handling, null-delimited processing for file lists to handle special characters safely, and proper argument separation using -- for grep and ripgrep commands.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from Git conflict markers, which constitutes an indirect prompt injection surface. The instructions effectively mitigate this risk by requiring human-in-the-loop verification for all resolutions and strictly forbidding the agent from making autonomous decisions when the intended state is not clear from repository evidence.
[SAFE]: The skill emphasizes 'hard stops' and safety protocols throughout its workflow, requiring confirmation before mutating history or resolving conflicts, which prevents accidental data loss or unauthorized changes.

git-integrate