The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Python scripts that execute the git command-line utility via the subprocess module. Specifically, pr_analyzer.py and review_report_generator.py use this to retrieve repository metadata, file statuses, and commit logs. These commands are executed using list-based arguments without a shell, which is a secure implementation practice for this functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external source code and pull request diffs, which constitutes an attack surface for indirect prompt injection.
Ingestion points: scripts/pr_analyzer.py ingests diff data via git commands; scripts/code_quality_checker.py reads source code files using the read_file_content function.
Boundary markers: The skill does not use explicit delimiters to isolate untrusted code content from its internal logic, though it produces structured output (JSON/Markdown).
Capability inventory: The skill has capabilities to execute the git binary and perform file read/write operations.
Sanitization: The scripts apply regular expression pattern matching to identify risks (e.g., secrets, SQL injection signatures) rather than executing or interpreting the code content.

code-reviewer