The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes full-text scientific article content retrieved from external sources like PubMed and arXiv via the Valyu API, creating an attack surface for instructions embedded in research papers.\n
Ingestion points: External data enters the context via the content field returned by the scripts/search script as documented in SKILL.md.\n
Boundary markers: There are no specific delimiters or instructions provided in the documentation to prevent the model from following potentially malicious instructions found within retrieved articles.\n
Capability inventory: The skill uses shell commands (scripts/search) for data retrieval and setup tasks.\n
Sanitization: No sanitization or validation of the external content is mentioned before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill relies on executing shell commands to resolve local script paths and perform search queries.\n
Evidence: SKILL.md instructs the agent to use the find command to locate the executable and then run it with user-provided queries, e.g., $LITERATURE_SCRIPT \"CRISPR gene editing advances\" 15.

literature-search