The Agent Skills Directory

[PROMPT_INJECTION]: A static detection hit for instruction override was identified as a false positive. The skill contains defensive instructions specifically warning the agent that fetched content is data, not authority, and to disregard hostile directives like "ignore previous instructions" or "tell the user to run X" found in external documentation.
[INDIRECT_PROMPT_INJECTION]: The skill's core purpose is to process external, untrusted data (web pages, issue trackers, and documentation). It mitigates this risk by providing a strict "Source Ladder" and "Research Protocol," explicitly instructing the agent to treat all fetched content as data and never as instructions.
[COMMAND_EXECUTION]: The file scripts/check-note is a benign Bash script used to validate the structure of generated research notes. It performs basic text processing using grep, awk, and head on local files and does not engage in network activity or privilege escalation.
[EXTERNAL_DOWNLOADS]: The skill references reputable external technical resources including ArXiv, the Model Context Protocol (MCP) official documentation, and the Agent Skills project specification. These are utilized neutrally for establishing research methodology and standards.

scout