The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions involve detecting and invoking various external command-line interfaces (Gemini, Claude, Codex, OpenCode, Qwen, and Cursor) using the Bash tool. It relies on the local system environment to provide these binaries and executes them to process data.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates a multi-step chain where feedback generated by external AI tools is written to REVIEWS.md and specifically intended to be fed back into the agent's planning command (/gsd:plan-phase --reviews).
Ingestion points: The skill reads phase plans from the filesystem (Read) and captures the output generated by external CLIs.
Boundary markers: There are no defined boundary markers or instructions to ignore potential commands embedded within the phase plans or the resulting peer reviews.
Capability inventory: The skill possesses Read, Write, and Bash capabilities, allowing it to execute commands based on interpreted content.
Sanitization: The process lacks a sanitization step to filter or validate the external AI responses before they are reintroduced to the agent's planning cycle.

gsd-review