The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external web sources, which introduces a surface for indirect prompt injection.
Ingestion points: External websites, official documentation, and community forums are crawled during Phase 2.
Boundary markers: The skill requires structured output files but does not specify the use of delimiters to segregate untrusted external content from internal instructions.
Capability inventory: The agent can write files to user-defined paths and execute system commands via bash for statistical analysis.
Sanitization: There is no explicit requirement for the agent to sanitize or filter retrieved content before it is processed.- [DYNAMIC_EXECUTION]: The methodology suggests using bash tools to perform statistical analysis on research data. This dynamic command execution is constrained to the primary purpose of technical research.- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs network requests to collect research data from authoritative sources. This behavior is documented as a core function of the skill and targets reputable information repositories.

deep-researcher