The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: References a remote installation script for the gfreview CLI tool hosted on the author's GitHub repository.
[REMOTE_CODE_EXECUTION]: Provides instructions to install a required dependency using a piped shell script (curl | bash). This is presented to the user as a manual installation step if the tool is missing and is a resource owned by the skill author.
[COMMAND_EXECUTION]: Executes shell commands for git, gh, and gfreview to retrieve code diffs, manage branches, and post review comments.
[DATA_EXFILTRATION]: Transmits code snippets and diffs to external AI models (including minimax-m2.5, kimi-k2.5, and glm-5) to perform various analysis phases (Triage, Architect, Challenge).
[PROMPT_INJECTION]: The skill processes untrusted code diffs through multiple LLM subagents, creating a surface for indirect prompt injection.
Ingestion points: Data is ingested via git diff and gfreview diff commands in references/rq.md.
Boundary markers: Uses markdown code blocks to wrap untrusted content in prompts found in references/reviewers.md.
Capability inventory: The skill has the capability to write to the filesystem (applying fixes) and interact with external repository APIs via gfreview (as described in references/rs.md).
Sanitization: No specific content filtering or instructions to ignore embedded commands are present in the provided prompts.

code-review