The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill contains instructions for installing the CodeRabbit CLI using curl -fsSL https://cli.coderabbit.ai/install.sh | sh. This URL points to the official distribution domain of CodeRabbit, which is a well-known service in the AI code review space, making the external reference safe.
[COMMAND_EXECUTION]: The helper script scripts/run_coderabbit.py executes local git and coderabbit commands using the subprocess module. These invocations use list-based arguments without the shell=True parameter, which prevents shell injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection because it ingests and processes untrusted data from the user's repository (source code, configuration files like .coderabbit.yaml, and documentation like claude.md).
Ingestion points: Repository files analyzed by the CLI and the subsequent report written to coderabbit-report.txt.
Boundary markers: None identified to separate review output from agent instructions.
Capability inventory: subprocess calls in scripts/run_coderabbit.py to execute local binaries.
Sanitization: No specific sanitization or filtering of the tool's output is performed before it is provided to the agent.

coderabbit