The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local bash script to perform literature metadata retrieval.
Evidence: The scripts/search.sh file is invoked with agent-generated arguments to query the Qinyan Academic OpenAPI via curl.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied search queries which are interpolated into shell command arguments.
1. Ingestion points: User-provided bibliographic information such as titles, authors, and keywords as described in the SKILL.md workflow.
1. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when constructing the search payload.
1. Capability inventory: Execution of local bash scripts and network communication via curl as defined in scripts/search.sh.
1. Sanitization: The scripts/search.sh file includes a validation step using python3 to verify that the search payload is a valid JSON object before the network request is initiated, providing a layer of structural integrity.

qinyan-citation