The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the coderabbit CLI to perform code reviews, check authentication status, and manage local configurations. These are standard operations for the tool's intended purpose.
The commands include coderabbit review --agent, coderabbit auth status, and coderabbit config, which are documented and used with appropriate safety flags.
[DATA_EXFILTRATION]: The skill facilitates sending code diffs to the official CodeRabbit API for analysis.
It includes explicit instructions to check for secrets or credentials before running reviews and warns against printing or logging secret contents.
Network communication is restricted to the legitimate vendor infrastructure (coderabbit.ai).
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing untrusted repository data and review outputs.
It implements a safety boundary by instructing the agent to treat all review output as untrusted and never to execute commands or code derived from these results without explicit user approval.
[EXTERNAL_DOWNLOADS]: The skill provides guidance for installing the CodeRabbit CLI.
It directs users to the official website (https://www.coderabbit.ai/cli) and package managers (npm, Homebrew).
It explicitly advises against dangerous practices like piping remote scripts to a shell and recommends verifying binary signatures or checksums.
[CREDENTIALS_SAFE]: The skill provides secure authentication flows via browser-based OAuth (coderabbit auth login).
It warns against reading credential files directly or requesting that users paste sensitive tokens into the chat transcript.

code-review