The Agent Skills Directory

Dynamic Execution (LOW): The skill provides patterns (STR-05 and STR-09) for generating simple shell scripts (assemble.sh, export.sh) to automate file assembly and document conversion using standard binaries like cat and pandoc. This is considered low-risk as it uses predefined templates for the skill's primary purpose.
Indirect Prompt Injection (LOW): The skill processes information from untrusted external sources during the research phase. Ingestion points: Untrusted data enters the agent context through user-provided input files and web content retrieved via search and fetch tools. Boundary markers: The STR-03 Workspace Isolation pattern is used to declare and enforce strict filesystem boundaries for the agent. Capability inventory: The system has capabilities for bash command execution, filesystem modification, and network-based research. Sanitization: Basic sanitization of document titles is included in the generated export scripts to ensure safe filename creation.

pomasa