The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates complex workflows using tmux, git, and the claude CLI. It specifically encourages the use of the --dangerously-skip-permissions flag, which allows the underlying AI agent to execute arbitrary shell commands and perform file modifications without any manual approval or oversight. This configuration significantly increases the risk of the agent performing unintended or destructive actions if it receives malicious instructions.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of instructions and commands generated by a remote large language model by providing an automation wrapper for the claude-code utility.
[PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection during its multi-model code review phase (Step 6). It processes untrusted data from git diffs and passes it directly to LLMs without isolation or sanitization. 1. Ingestion points: Git diff data obtained via gh pr diff in SKILL.md. 2. Boundary markers: None identified in the provided review prompts. 3. Capability inventory: Full shell access through tmux, git for repository manipulation, and the claude CLI for file edits. 4. Sanitization: No sanitization, escaping, or validation of the diff content is performed before it is processed by the reviewer models.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the @anthropic-ai/claude-code package. This is a legitimate utility provided by Anthropic, which is a recognized and trusted organization.

claude-skill