The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses shell pipes to pass curl output to python3. Although automated detectors flag this as a risk, the Python code is hardcoded and used only for parsing the API's JSON response. The remote content is processed as data and never executed.
[EXTERNAL_DOWNLOADS]: Fetches research data from api.crossref.org, a well-known academic service. This activity is legitimate and safe within the context of academic search.
[COMMAND_EXECUTION]: Employs standard CLI tools (curl and python3) for data retrieval and formatting. These operations are scoped to the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from the CrossRef API, which could theoretically contain malicious instructions.
Ingestion points: Academic paper titles, authors, and abstracts retrieved from the API.
Boundary markers: None identified.
Capability inventory: Ability to perform network requests and execute Python scripts.
Sanitization: Uses json.load() to parse responses, which ensures the data is treated as structured objects rather than instructions.

crossref-search