The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool for Git operations, diff analysis, and applying 'Auto-Fix' suggestions. It also includes functional installation scripts (hooks/install.sh, hooks/install.ps1) and a Git pre-push hook which execute commands to manage the local Git configuration.
[EXTERNAL_DOWNLOADS]: The skill README and documentation reference the author's GitHub repository (github.com/Wubabalala/claude-skills) for project-specific updates and issue tracking.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8) because it reads and follows instructions from .claude/review-checklist.md and architecture-traps.md files found within the target repository. An attacker who can commit to the repository could include malicious instructions in these files to influence the agent's review or bypass checks.
Ingestion points: architecture-traps.md, .claude/review-checklist.md, and repository code files.
Boundary markers: The skill lacks explicit system-level instructions or delimiters to isolate the agent from potentially malicious prompt injections within these project files.
Capability inventory: The agent has access to Bash, Write, Edit, Read, Grep, and Glob tools, providing a significant blast radius if an injection is successful.
Sanitization: No validation or sanitization of the logic inside the ingested configuration files is performed.

code-review