The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts pr_analyzer.py and review_report_generator.py use subprocess.run to execute the git binary and sub-scripts for analysis. While the implementations use argument lists to avoid shell injection, the skill possesses the capability to spawn external processes based on user-provided repository paths and branch names, which is a necessary function for its purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from git diffs and source code files, creating a potential surface for indirect prompt injection.
Ingestion points: Files are read in scripts/code_quality_checker.py via standard file operations, and git diffs are captured in scripts/pr_analyzer.py through command output.
Boundary markers: Analysis results are aggregated into reports in scripts/review_report_generator.py without explicit delimiters or instructions to the AI agent to ignore commands potentially embedded in the analyzed code.
Capability inventory: The skill can execute local commands via subprocess.run and write results to the local filesystem.
Sanitization: The ingested data is processed for metrics via regex but is not sanitized to remove potential prompt injection sequences before being included in review reports.

code-reviewer