The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands (git rev-parse, git diff, git log) using a user-provided 'fixed-point' value. This poses a command injection risk if the input contains shell metacharacters and is not correctly escaped by the agent platform.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of untrusted repository content. Ingestion points include git diff output, commit logs, and documentation/specification files from the repository. No specific boundary markers or encapsulation techniques are mentioned to separate untrusted data from the sub-agent instructions. The skill utilizes shell-based git tools and creates new sub-agent contexts to process analysis results. The instructions lack any requirement for sanitizing the repository-sourced data before its inclusion in prompts.

code-review