The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external web searches and user-provided arguments through the /research-lit and /idea-creator sub-skills. There are no explicit boundary markers or sanitization steps mentioned to isolate this external content, which is a standard surface for indirect prompt injection common in research agents.
Ingestion points: Data enters the context via WebSearch, WebFetch, and the $ARGUMENTS variable.
Boundary markers: Instructions do not specify the use of delimiters or 'ignore' instructions for external content.
Capability inventory: The skill uses Bash(*), Write, Edit, and triggers other Agent skills.
Sanitization: No explicit validation or escaping of external content is performed.
[COMMAND_EXECUTION]: The skill is configured to use the Bash tool for project orchestration. It includes an AUTO_PROCEED setting that allows the pipeline to move through phases (including file modifications and tool calls) without waiting for human intervention if the user does not respond, which reduces oversight of autonomously generated actions.

idea-discovery-robot