The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides a command to download an installation script from https://cli.coderabbit.ai/install.sh. This source belongs to CodeRabbit, a well-known service provider in the AI-powered code review space.- [REMOTE_CODE_EXECUTION]: The installation instructions use a pipe to a shell (curl | sh) to execute the remote script. This is a common and accepted installation pattern for developer tools from established vendors.- [COMMAND_EXECUTION]: The skill facilitates the execution of various coderabbit CLI commands for local reviews, authentication, and managing cloud-based coding tasks.- [INDIRECT_PROMPT_INJECTION]: The skill provides an interface to a service that processes untrusted data (source code and pull request comments).
Ingestion points: Code diffs from local repositories or remote branches, and comments/threads from GitHub pull requests.
Boundary markers: The documentation recommends using the --agent flag for structured JSON output and describes the event-based schema to help the agent distinguish between tool status and review content.
Capability inventory: The tool can read and upload code, write commits/pushes to the repository (via code push or Autofix), and manage authentication state.
Sanitization: The skill explicitly instructs the agent and user to validate all findings before applying fixes, treats unverified findings separately, and emphasizes that the user must explicitly approve actions that involve billing or code modification.- [PRIVILEGE_ESCALATION]: The references/allowlist.md file suggests configuration patterns to auto-approve certain CLI commands in the agent's execution environment. It mitigates risk by categorizing commands and recommending manual approval for high-risk operations like binary updates and authentication.

coderabbit