The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell utilities like ls, grep, jq, and cat, alongside a project-specific CLI br, to verify task status and artifact integrity. It also triggers project-specific build and test suites as mandatory quality gates.
[PROMPT_INJECTION]: The skill processes implementation artifacts (modified code and context documents) by passing them to subagents. While this involves processing potentially untrusted data (Indirect Prompt Injection surface), the workflow includes significant mitigations, such as a dedicated security specialist subagent and a mandatory human-in-the-loop verification phase (UAT).
[DATA_EXFILTRATION]: No unexpected network operations or data exfiltration patterns were detected. All file interactions and shell commands are scoped to the local repository and the project's management tools.

beo/reviewing