The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions describe executing a local shell script (scripts/setup-output-dir.sh) to create a timestamped directory structure and uses the Task tool to orchestrate parallel agent execution. It also utilizes the date command for timestamp generation.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a research aggregator that fetches data from the web using WebSearch and WebFetch. This creates a vulnerability surface where malicious instructions embedded in web content could potentially influence the downstream specialized agents. * Ingestion points: External market data and news are ingested via WebSearch and WebFetch tools as described in the research capabilities. * Boundary markers: The documentation does not specify the use of clear delimiters or instructions to isolate fetched content from agent instructions. * Capability inventory: The skill environment includes powerful tools such as Bash (shell execution), Write (file system access), and Task (sub-process orchestration). * Sanitization: There is no evidence of content sanitization or validation before the fetched data is passed to the specialized agents for analysis.

crypto-research