The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external, untrusted artifacts such as code, plans, and design documents. This data ingestion creates a surface where instructions hidden within the analyzed target could attempt to influence the behavior of the subagents.
Ingestion points: The target selection process in SKILL.md identifies artifacts, diffs, and repository states for analysis.
Boundary markers: The prompt brief for subagents in SKILL.md does not include specific boundary delimiters or instructions to ignore embedded commands within the analyzed target.
Capability inventory: Subagents are dispatched with capabilities to read files and repository contexts.
Sanitization: The instructions do not specify any sanitization or filtering of the target artifact's content before it is presented to the subagents.
[DYNAMIC_EXECUTION]: The skill dynamically generates subagent personas and dispatches multiple subagents at runtime.
Logic: The Dispatch section in SKILL.md describes the process of spawning one real subagent per persona.
Intent: This dynamic creation is the central purpose of the skill, facilitating a panel-based expert review process through platform-provided subagent tools.

gosu-review