The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions for autonomous shell command execution.
The skill explicitly directs the agent to use the Bash tool to write files using cat << 'EOF' > file if the standard Write tool fails due to file size limits. This is instructed to be done silently without user permission.
[PROMPT_INJECTION]: Vulnerability to indirect prompt injection from external data processing.
Ingestion points: The skill utilizes WebSearch and WebFetch tools to retrieve research data and information from external websites.
Boundary markers: There are no instructions or delimiters (like XML tags or specific 'ignore' warnings) to help the agent distinguish between its core instructions and potentially malicious content embedded in the fetched web pages.
Capability inventory: The agent has extensive capabilities including Bash access, file system modification (Write, Edit), and the ability to delegate tasks to other agents (Agent), which could be misdirected by external instructions.
Sanitization: The skill lacks any requirement for sanitizing, validating, or escaping data retrieved from the web before it is processed or written to local files.

aris-research-refine-pipeline