The Agent Skills Directory

Command Execution: The skill involves the execution of a local Python script (scripts/sample_size_calculator.py) to perform statistical calculations. In references/04-protocol-operations.md, the agent is instructed to construct and run a shell command using parameters such as --effect-size and --std-dev. These parameters are derived from user input or previously generated metadata. While this is necessary for the skill's purpose, it represents a high-capability behavior that could be susceptible to command injection if inputs are not properly sanitized before being passed to the shell.
Indirect Prompt Injection Surface: The skill features a modular architecture that ingests data from external sources and user-provided descriptions across multiple steps. This establishes an indirect prompt injection attack surface.
Ingestion points: Data enters the agent's context through user input in references/00-initialize-intervention.md and through ClinicalTrials.gov search results via an MCP server in references/01-research-protocols.md.
Boundary markers: The instructions do not explicitly specify the use of boundary markers or delimiters when interpolating this external data into the prompts for subsequent protocol generation steps.
Capability inventory: The skill can execute local scripts (scripts/sample_size_calculator.py), write files to the waypoints/ directory, and invoke external search tools via the MCP server.
Sanitization: There are no explicit instructions within the skill for the agent to sanitize or escape the content retrieved from external databases or user inputs before processing it in later stages.
External Package Installation: The SKILL.md file provides instructions for installing external Python dependencies (scipy, numpy) using pip. These are well-known, established libraries from standard registries and are required for the statistical functions of the skill.

clinical-trial-protocol-skill