The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a set of shell scripts (parallel-research, export-pdf) and system utilities like pandoc, jq, and ln to perform research activities and manage output files.
[REMOTE_CODE_EXECUTION]: Installation instructions include fetching the uv tool from a remote source via a shell script. Additionally, the skill uses uvx to dynamically download and execute Python dependencies at runtime for PDF conversion.
[PERSISTENCE]: The skill uses cron jobs to schedule future agent actions for research polling and modifies the user's shell profile (~/.bashrc) to persist API configurations.
[DYNAMIC_EXECUTION]: A Python script is dynamically generated and executed via a bash heredoc in the export-pdf utility to facilitate document processing using the pymupdf library.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process content from web searches and third-party API reports, creating a surface for indirect prompt injection.
Ingestion points: Fetches data from external web sources and markdown reports via the Parallel AI API.
Boundary markers: No specific delimiters or safety instructions are defined to separate ingested data from agent commands.
Capability inventory: Includes file system access, network communication, and execution of internal scripts.
Sanitization: External data is incorporated into research documents without dedicated validation or filtering routines.

research