The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and reference files direct the agent to execute several shell commands to gather context and perform reviews. These include standard developer tools like git (e.g., git diff, git status), gh (e.g., gh pr view, gh pr diff), and rg (ripgrep) for searching code.
[COMMAND_EXECUTION]: The review-panel-protocol.md file specifies the execution of a local binary located at <project-plugin-root>/bin/project-preflight. This is used as a "dispatch gate" to check for native agent availability before initiating a multi-agent review panel.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze untrusted data in the form of pull request diffs and source code files, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through commands like git diff, gh pr diff, and direct file reads of the project's source code.
Boundary markers: The instructions do not specify the use of clear boundary markers (such as XML tags or specific delimiters) to separate the untrusted code being reviewed from the agent's internal instructions.
Capability inventory: The agent has access to powerful CLI tools (git, gh) and the ability to execute a local project-preflight binary. The gh tool typically has permissions to modify pull requests, including adding comments or merging.
Sanitization: There are no explicit instructions for the agent to sanitize or escape the content of the diffs before processing them.

pr-review-toolkit