The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill possesses a significant attack surface by ingesting and processing untrusted external content.
Ingestion points: The pex ingest-md command reads local markdown files from a user-provided path into the system.
Boundary markers: No boundary markers or 'ignore' instructions are defined for the ingested data.
Capability inventory: The skill can execute the pex CLI, perform graph modifications (pex clear), and initiate LLM inference (phase6-infer).
Sanitization: There is no evidence of sanitization or safety filtering for content ingested from the markdown files before it is processed by the LLM.
Command Execution (MEDIUM): The skill relies on the execution of a local binary/script pex. While the source of the CLI is not specified, it is granted the ability to interact with the filesystem and external services like FalkorDB and Ollama.
Data Exposure (LOW): The pex ingest-md command allows for the reading of arbitrary directory paths. If exploited via a prompt injection or malicious input, this could be used to scan the local filesystem for sensitive data, though it is primarily designed for medical education data.

pex-cli