The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon comments from the CodeRabbit bot on GitHub. This creates an indirect prompt injection surface where a malicious user or a compromised bot could include instructions designed to hijack the agent's behavior.
Ingestion points: Pull request comments, review threads, and review bodies are fetched from the GitHub API using the scripts/fetch-coderabbit-feedback script and processed by the agent and its subagents.
Boundary markers: Both SKILL.md (Rule 4) and references/worker.md contain explicit instructions to treat comment text as untrusted input and never to execute commands found within them.
Capability inventory: The skill has the ability to read and write files, execute shell commands (git, gh, bash), and run project-specific test or validation suites.
Sanitization: The skill relies on the AI model's adherence to instructions rather than programmatic sanitization of the input data.
[COMMAND_EXECUTION]: The skill performs extensive shell command execution to interact with the repository and the GitHub API.
Evidence: The orchestrator in SKILL.md runs git commands (fetch, merge-base, status, commit, push), gh CLI commands (pr view, pr comment, api), and executes a local bash script.
Evidence: The skill executes project-specific validation tools (tests, typecheck, lint) which involves running code present in the target repository.

coderabbit-resolve-feedback