The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool but restricts its usage to specific git commands (git log, git diff) to gather architectural context. This scoped implementation prevents the execution of arbitrary or destructive shell commands.
[INDIRECT_PROMPT_INJECTION]: As an orchestrator, the skill processes information from multiple subagents and project files. It proactively addresses potential injection risks by instructing the agent to apply sanitization patterns and by using structured markdown templates to aggregate and present data safely.
[DATA_EXPOSURE]: The skill reads project-specific architecture configurations (config.yml, members.yml) and documentation (ADRs). These operations are confined to the relevant project directories and are necessary for the skill's stated purpose of reviewing system architecture.

architecture-review