The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/research.py executes external binaries (scholar and consensus) using subprocess.run and subprocess.Popen. It passes the user's search query as an argument to these commands. While it uses argument lists (preventing direct shell injection), the execution of local binaries based on user input remains a sensitive capability.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection from external academic metadata.
Ingestion points: Metadata including 'takeaways' and 'titles' are fetched from Google Scholar, Consensus, and CrossRef APIs via scripts/research.py.
Boundary markers: Absent. The instructions in SKILL.md do not specify delimiters or instructions to ignore potential commands embedded in paper abstracts.
Capability inventory: The skill has the ability to execute shell commands via subprocess in scripts/research.py.
Sanitization: Absent. The script passes external content directly to the agent's context for presentation to the user.
[EXTERNAL_DOWNLOADS]: The script fetches paper metadata from api.crossref.org. CrossRef is a well-known academic service used for DOI resolution, and these requests are considered standard functionality for a research tool.

research