The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection due to its core functionality of processing external data.
Ingestion points: According to the 'research brief' in SKILL.md, the agent is instructed to investigate official documentation, package source code (e.g., node_modules, vendor), specifications, and first-party APIs.
Boundary markers: The instructions do not specify any boundary markers or delimiters to help the sub-agent distinguish between the user's research task and potential instructions embedded within the primary sources being read.
Capability inventory: The skill dispatches sub-agents (Explore and general-purpose) which have the capability to perform web fetches, call other tools, and write files to the repository.
Sanitization: There are no instructions for sanitizing or filtering external content to prevent the agent from executing instructions found in the documentation or code it processes.

research