The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Git commands such as git log, git diff, and git merge-tree via a Bash script to perform repository analysis. These operations are limited to read-only and dry-run tasks, and the skill contains explicit instructions prohibiting automated commits, merges, or pushes.
[PROMPT_INJECTION]: The skill processes untrusted data from the repository's commit history and file structure, which creates a surface for indirect prompt injection.
Ingestion points: Commit messages and file names are read from the repository into the agent's context through git log and git diff output in scripts/pre-merge-check.sh.
Boundary markers: The skill formats the analysis results as a structured JSON object, providing a clear boundary for the agent between metadata and content.
Capability inventory: The skill is granted access to the Bash, Read, Grep, and Glob tools.
Sanitization: The Bash script uses awk to escape double quotes in commit messages before including them in the JSON output to prevent schema confusion and formatting errors.

merge-prep