The Agent Skills Directory

[SAFE]: The skill defines a role for a senior protocol engineer to perform safety-critical code reviews. The instructions are focused on technical analysis of blockchain protocol changes and follow standard best practices for identifying bugs in distributed systems.
[COMMAND_EXECUTION]: The prompt directs the agent to perform git diff commands to examine code changes. This is a legitimate and expected use of development tools for the stated purpose of code review and does not involve arbitrary or high-risk command execution.
[PROMPT_INJECTION]: The skill processes untrusted code changes via git diff output (Ingestion point: SKILL.md). It does not provide specific delimiters to separate code from instructions (Boundary markers: absent) or sanitization steps for the input data (Sanitization: absent). The agent's capabilities are limited to reading and analyzing the diff (Capability inventory: git diff). While this presents a surface for indirect prompt injection, the skill itself contains no malicious instructions or capabilities for data exfiltration or persistence.

cmd-chain-halt-code-reviewer