The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from repository-local files to configure its behavior, which could be exploited if a repository contains malicious instructions.
Ingestion points: The skill reads production/review-mode.txt (Phase 0) and other project documentation like AGENTS.md and CLAUDE.md to determine its review mode and instructions.
Boundary markers: No specific delimiters or safety warnings are implemented for these local file reads.
Capability inventory: The skill has access to Bash, Write, Edit, and Task (subagent spawning) tools.
Sanitization: The skill does not explicitly sanitize the content retrieved from these local configuration files.
[COMMAND_EXECUTION]: The skill utilizes the Bash and Task tools for performance profiling and to delegate work to subagents. These capabilities are aligned with the skill's primary purpose of orchestrating a technical polish team.

team-polish