The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes paper abstracts and titles from PubMed via the pubmed_search.py script. It lacks explicit instructions or boundary markers to differentiate between the retrieved research content and the agent's instructions, creating a surface for indirect prompt injection if an abstract contains malicious directives.
Ingestion points: PubMed abstracts and titles retrieved by pubmed_search.py and processed by the agent in SKILL.md (Step 3).
Boundary markers: Absent in the synthesis and analysis instructions.
Capability inventory: Shell command execution via bash for running the search script; file system write access for saving markdown summaries.
Sanitization: None provided for the retrieved external text before it is analyzed by the language model.
[COMMAND_EXECUTION]: The skill uses shell commands to execute the local pubmed_search.py script, passing user-provided search terms as arguments. While functional, this relies on the agent's ability to safely format arguments to prevent command injection at the shell level.

medical-research